NeuralWired’s Technology section covers the developments reshaping how the world builds, deploys, and regulates digital innovation. We report daily on the stories driving global conversation in artificial intelligence, big technology companies, startups and venture funding, cybersecurity, consumer gadgets and devices, and blockchain and cryptocurrency.
Our technology coverage goes beyond product announcements. When a major AI model launches, we explain what it can actually do and where its claims are overstated. When a startup raises a large funding round, we look at whether the business behind it can sustain that valuation. When a cybersecurity breach hits the news, we explain who is affected and what comes next, not just what happened. Each article is built from original research into primary sources, including company statements, technical documentation, regulatory filings, and verified data, and is written by our editorial team rather than generated automatically.
Readers come to this section for daily updates on the technology stories that matter globally, from shifts inside major technology companies to emerging tools changing how people work, communicate, and build. Whether you are a founder, an investor, an engineer, or simply someone trying to understand where technology is heading next, NeuralWired’s Technology coverage is built to keep you informed without wasting your time on hype.
Compliance-as-Code: How Developers Meet the EU AI Act
Policies / Developer Focus
Compliance-as-Code: How Developers Meet the EU AI Act
By the NeuralWired Staff · Updated July 24, 2026 · 9 min read
A developer at a mid-size fintech company got a Slack message on a Tuesday: legal needed to know, by Friday, whether the new fraud-detection model complied with the EU AI Act. The honest answer was nobody knew, because the compliance policy lived in a PDF last opened eight months earlier. That gap, between a static document and a system that changes every sprint, is exactly what AI compliance as code is built to close.
Engineering teams are no longer waiting for legal to translate regulation into requirements after the fact. They’re writing the rules directly into the pipeline, as version-controlled policy files that run automatically on every pull request, every build, every deploy, and that can stop a release cold if something violates the rule. This is compliance-as-code, and it’s moving from a niche DevSecOps practice into the default way teams handle AI governance in 2026.
Here’s the uncomfortable number that should be driving this conversation more than any regulation: according to Vanta data reported in July 2026, roughly 70% of companies now have shadow AI somewhere in their environment, meaning AI tools with access to company data that were never vetted or approved. Shadow IT overall grew 36% year over year, and organizations discovered an average of about 140 unapproved tools within 90 days of scanning their environment. Only 2% of shadow IT vendors ever go through a security review at all. (Source: KESQ/Stacker, July 22, 2026)
A quarterly compliance review can’t see any of that. It’s a snapshot of a system that’s already moved on by the time the report gets written. That mismatch, static review versus continuously changing systems, is the actual argument for compliance-as-code. It’s not a regulatory checkbox exercise. It’s the only way to see what’s happening in real time.
The core shift: Compliance moves from something legal owns and checks once a quarter, to something the platform team owns and enforces on every single change, the same way failing unit tests block a merge today.
What Compliance-as-Code Actually Means
Strip away the buzzwords and the practice is simple. Instead of a written policy document that a human is supposed to remember to consult, the rule gets written in a machine-readable, declarative language, most commonly Rego, HashiCorp Sentinel, YAML, or CEL, and that rule gets checked automatically at defined points in the software lifecycle: a pull request, a CI build, a Terraform plan, a Kubernetes admission event, or an AI agent’s proposed action.
If the rule is violated, the pipeline fails. Not a warning. Not a note for someone to review later. The build stops.
For AI-specific obligations, some teams are now storing a single compliance.yaml file at the root of the model repository that auto-generates the technical documentation regulators expect under the EU AI Act’s Article 11 and Article 12 record-keeping requirements. That’s a practitioner pattern worth watching, not yet an industry standard, so treat it as an emerging convention rather than something auditors will universally recognize.
A simple example
A policy blocking a high-risk model from deploying without a documented human-oversight mechanism might look something like this in Rego:
package ai.deployment
deny[msg] {
input.model.risk_tier == “high”
not input.model.human_oversight_documented
msg := “High-risk model missing required human oversight documentation”
}
That’s it. No PDF. No email chain. The pipeline reads it, checks it, and either lets the deploy through or doesn’t.
The EU AI Act Deadline Forcing the Shift
The EU AI Act isn’t a distant threat anymore. Its prohibited-practices provisions have been enforceable since February 2, 2025. General-purpose AI model obligations kicked in on August 2, 2025 for new models. And the next major wave, obligations for high-risk systems, has a baseline date of August 2, 2026, with the EU’s Digital Omnibus proposal actively negotiating whether some sub-provisions shift into a 2027 window. The trilogue process on that shift is still active as of mid-2026, so treat “full compliance by August 2026” as directionally accurate but not settled in every detail. (Source: Holland & Knight, April 28, 2026)
Penalties for prohibited practices reach €35 million or 7% of global annual turnover, whichever is higher. And critically, this isn’t an EU-companies-only problem. Any provider, deployer, importer, or distributor whose system touches EU residents is in scope, regardless of where the company is headquartered. If your app has European users, this applies to you.
U.S. state law is moving just as fast, if less dramatically. Colorado’s original AI Act was frozen by a federal court and then repealed after litigation from Elon Musk’s xAI, replaced by SB 26-189, which sets new employer obligations. NeuralWired covered what that means for employers here. The pattern across jurisdictions is the same: rules that shift faster than any static document can track.
Inside the Stack: OPA, Gatekeeper, and the Tooling
The engine underneath most of this is Open Policy Agent, or OPA, a Cloud Native Computing Foundation graduated project, meaning it’s cleared the foundation’s highest maturity bar. OPA policies are version-controlled and reviewed through the same Git workflows developers already use for application code, which is precisely why adoption has been so fast: there’s no new mental model to learn. (Source: Wiz Academy)
According to a DevOpsTales analysis citing Snyk’s State of DevSecOps report and the SANS DevSecOps Survey, policy as code now has 71% overall enterprise adoption as a practice, with OPA specifically cited by 68% of enterprises as the leading tool. That’s not a niche technique anymore. It’s the default. (Source: DevOpsTales, Sept 2025)
Gatekeeper, the Kubernetes-native project that extends OPA, keeps shipping too, which matters because it shows this tooling is still actively maturing rather than sitting still. Gatekeeper reached general availability for CEL-based validating admission policies in version 3.18, management for those policies went to beta in 3.20, and in version 3.22 (February 2026) a scoped enforcement feature was enabled by default. (Source: Spacelift)
Quick comparison
Tool
Best for
Maturity signal
Open Policy Agent (OPA)
General-purpose policy across infra, APIs, CI/CD
CNCF graduated; 68% cited enterprise adoption
Gatekeeper
Kubernetes admission control
~4,200 GitHub stars; CEL-based policies GA in v3.18
HashiCorp Sentinel
Terraform plan enforcement
Embedded natively in Terraform Cloud/Enterprise
GitHub stars are one adoption signal among several, not a definitive market-share measure.
The Cost of Getting This Wrong
Gartner estimates global spend on AI governance platforms will reach $492 million in 2026, and cross $1 billion by 2030. Organizations that actually use these platforms are 3.4 times more likely to achieve high effectiveness in AI governance than those that don’t. (Source: Gartner, Feb 17, 2026)
The market’s growth also earned Gartner’s first-ever Magic Quadrant for AI Governance Platforms, published June 16, 2026, a formal signal that analysts now treat this as its own distinct category rather than a subset of general GRC tooling. (Source: Credo AI, citing Gartner)
On the downside-risk ledger, non-compliance carries costs well beyond the headline fine. Elevate Consult’s compiled 2026 analysis puts customer trust and revenue loss at 15% to 30% of affected revenue streams, legal costs between £500,000 and £5 million, operational disruption lasting three to twelve months, and insurance premium increases of 25% to 50%. (Source: Elevate Consult, May 28, 2026)
“Enterprises are treating AI agent governance as binary, either locked down or fully trusted, and that is the root cause of failure.”
— Shiva Varma, Senior Director Analyst, Gartner
Varma’s team backs that up with a genuinely striking prediction: by 2027, Gartner expects 40% of enterprises to demote or decommission autonomous AI agents entirely, because governance gaps only surface after something has already gone wrong in production. (Source: Gartner, May 26, 2026) NeuralWired’s earlier reporting on why 70% of AI agent deployments fail lines up with that pattern almost exactly.
“Cybersecurity leaders must identify both sanctioned and unsanctioned AI agents, enforce robust controls for each, and develop incident response playbooks to address potential risks.”
— Alex Michaels, Director Analyst, Gartner
And the human bottleneck driving all of this toward automation has a number attached too: 61% of compliance teams report regulatory complexity and resource fatigue, according to Amanda Carty, General Manager of Compliance Solutions at Diligent (a vendor in this space, worth noting as you weigh her framing). (Source: Governance Intelligence)
Automated Doesn’t Mean Correct
Here’s the pushback worth sitting with: a green checkmark from an automated policy check is not automatically more trustworthy than a signed PDF, if nobody is independently reviewing the logic of the policy itself.
Compliance-as-code inherits every known failure mode of static analysis tooling. A study of financial-services engineers found that out of 200 initial security alerts, only about 40 turned out to be legitimate, and developers reported spending roughly 5% of their working time just triaging false positives. Applied to AI compliance specifically, this cuts both ways: a rule written too strictly blocks legitimate model updates, and a rule written too loosely creates exactly the false assurance a checkbox audit already gave everyone. (Source: arXiv, Secure Software Engineering in Financial Services)
There’s also a structural incentive problem, not just a technical one. Research on third-party AI safety-framework reviews found that auditors face real pressure, client retention, and conflicts of interest when the same firm sells consulting alongside auditing, that can produce compliance results which look clean but aren’t. The same logic applies the moment an internally-written Rego policy becomes the artifact regulators or customers are asked to trust. (Source: arXiv 2505.01643)
Separately, even mature tooling doesn’t fix a leadership problem. BlackFog’s 2026 survey found 69% of C-suite executives are effectively tolerating unsanctioned AI tool use, prioritizing speed over governance.
“The efficiency gains and personnel cost savings are too large to ignore, and override any security concerns.”
— Darren Williams, Founder and CEO, BlackFog
Our read: a CI gate that leadership can override whenever a launch is at risk isn’t really a gate. The technical fix and the incentive fix have to land together, or the pipeline just becomes a slower version of the PDF nobody read.
One more voice worth including here, because it reframes the problem usefully for the developer audience specifically:
“Put yourself in the shoes of a software developer: you have a question and your organization hasn’t put in place the tools to answer that question. Instead, you use your phone to browse an unapproved GenAI website and get your question answered promptly.”
— Doug Ross, CTO and GenAI Lead, Capgemini
In other words, shadow AI usually isn’t a discipline failure by developers. It’s a tooling gap left open by employers. Compliance-as-code is only a real fix if it also gives developers a faster, sanctioned path to the answer they need, not just a faster way to say no.
What to Do This Quarter
The tools here are not the hard part. OPA, Conftest, Gatekeeper, and Kyverno are open-source, mature, and free. The actual lift is organizational.
Decide who owns policy authorship. Security, platform, or legal-translated-by-platform. Ambiguity here is the single biggest reason compliance-as-code initiatives stall.
Turn logging into blocking. A policy that only logs a violation isn’t compliance-as-code. It’s a slightly fancier audit trail. The gate has to actually fail the build.
Tier your AI agents by autonomy, not by a single locked-or-trusted switch. Gartner’s own data points to this as the difference between governance that works and governance that fails in production.
Build the audit trail once, reuse it everywhere. Teams writing policy-as-code now are finding it satisfies SOC 2, ISO 42001, and GDPR evidence requirements at the same time as the EU AI Act, a genuine “test once, satisfy many” pattern.
Reality check on timing: Ranking on page one of Google within two to three days through on-page content alone isn’t something any legitimate SEO practice can guarantee, no matter how well a piece is optimized. What is realistic in that window is getting cited inside AI Overviews and answer engines like ChatGPT, Perplexity, and Gemini, because those systems pull from freshly published, clearly sourced, well-structured content much faster than traditional organic rankings move. That’s the actual near-term win this piece is built for.
FAQ
What is compliance as code?
Compliance as code is the practice of writing regulatory and security requirements as executable, version-controlled policy files that run automatically in CI/CD pipelines, blocking non-compliant deployments before they reach production instead of flagging them after the fact.
What is policy as code?
Policy as code means defining organizational, security, or compliance rules in a declarative language like Rego, then enforcing them automatically across infrastructure, applications, and CI/CD workflows, replacing manual policy review with automated, testable checks.
When does the EU AI Act take full effect?
Most remaining provisions, including high-risk system obligations, carry a baseline date of August 2, 2026, though the Digital Omnibus proposal is actively negotiating moving some sub-provisions to 2027. Prohibited practices have been banned since February 2, 2025.
What is Open Policy Agent used for?
Open Policy Agent is an open-source, CNCF-graduated policy engine that lets teams enforce rules across Kubernetes admission control, Terraform plans, APIs, and CI/CD pipelines, using a Git-based workflow so policies get reviewed and tested like application code.
How much does non-compliance with AI regulation cost?
Non-compliance can trigger EU AI Act fines up to €35 million or 7% of global annual turnover, plus indirect costs including 15% to 30% customer trust and revenue loss, legal costs of £500,000 to £5 million, and insurance premium increases of 25% to 50%.
What is shadow AI?
Shadow AI refers to AI tools employees use at work without formal IT or security approval. As of mid-2026, Vanta data shows roughly 70% of companies have some shadow AI present in their environment, most of it never vetted or reviewed.
Where This Goes Next
What changes here isn’t just how audits get done. It’s who owns compliance day to day. That ownership is shifting from legal, working off a document nobody updates, to platform engineers, working off a policy file that runs on every single commit. Developers will start encountering Rego and OPA failures in CI the same way they already encounter a failing unit test today, not as a special event, just as part of the job.
Watch three things over the next six to eighteen months: whether the EU’s Digital Omnibus trilogue actually pushes high-risk obligations into 2027 or holds the August 2026 line; whether a harmonized compliance.yaml-style schema for AI Act technical documentation gets formal recognition, since right now teams are building against an unofficial standard; and whether Gartner’s 40%-by-2027 agent-decommissioning prediction actually plays out, which would be the clearest signal yet that governance gaps, not model capability, are the real ceiling on enterprise AI agent adoption.
None of this is a purchasing decision. OPA is free. Gatekeeper is free. The teams that get ahead of the August 2026 deadline are the ones who decide, now, who owns the policy and whether the gate is actually allowed to say no.
Want this kind of analysis in your inbox? Subscribe to The Neural Loop at neuralwired.com/newsletter for weekly breakdowns of the regulation, tooling, and incidents shaping how engineering teams actually build with AI.
SpaceX Buys Cursor for $60B: Inside the AI Coding Tools War
Amazon is killing Q Developer. Google is retiring Gemini CLI. And SpaceX just bought the market leader in AI coding tools for more money than most countries’ GDP. Here’s what actually happened, and what you need to do about it.
If you picked an AI coding tool eighteen months ago, there’s a decent chance it doesn’t exist anymore, or won’t by next year. That’s the real story behind the 2026 AI coding tools shakeup, and it’s a lot messier than the tidy “Big Tech is consolidating” headline suggests.
Three products anchor this story: GitHub Copilot, Amazon Q Developer, and Gemini Code Assist. Only one of them is actually thriving. The other two are being shut down by their own parent companies. And the biggest deal of the year isn’t a tech giant tightening its grip. It’s a rocket company buying the market leader outright.
This Isn’t Consolidation. It’s a Demolition.
The comfortable narrative goes something like this: Microsoft, Google, and Amazon are quietly locking down the AI coding tools market, and independent players don’t stand a chance. As of July 2026, that story is only half right.
What’s actually happening is stranger. Amazon has now discontinued two coding assistants in under two years. Google is sunsetting the free version of its own command-line agent just months after launching it. And the most dramatic move in the entire category came from outside it entirely: SpaceX, fresh off a record-setting IPO, wrote a $60 billion check for Cursor’s parent company, Anysphere.
Every major player now sits inside Microsoft, Google, OpenAI, or SpaceX and xAI. Anthropic’s Claude Code is the one notable holdout, though even Anthropic carries investment from Amazon and Google. The independent era of AI coding tools, the one where Cursor, Windsurf, and standalone agent CLIs competed on their own terms, lasted roughly three years before folding into the platforms that fund the underlying models.
Amazon Q Developer: Dead in Under Two Years
Amazon Q Developer isn’t being folded into a bigger platform. It’s being retired, full stop. AWS confirmed on its official DevOps blog that Q Developer’s IDE plugins and paid subscriptions reach end of support on April 30, 2027, and new signups were already blocked as of May 15, 2026.
The replacement is Kiro, a standalone spec-driven agentic IDE built on Code OSS, the same open-source foundation as VS Code. AWS unveiled a Kiro Pro Max tier at $100 a month and a native iOS app at its June 2026 Summit in New York, and previewed a Kiro Autonomous Agent capable of running independently for days at a time. Early traction looks real: Kiro pulled in 250,000 users in its first three months.
Here’s the part that should worry anyone who’s been burned by this before: Q Developer was already a successor product. It replaced CodeWhisperer, which Amazon discontinued as a standalone tool in late 2025. That’s Amazon’s second coding-assistant sunset in under two years, and if you’re a platform architect who bet on either product, you’ve now migrated twice.
The pattern to watch: Amazon isn’t struggling to build AI coding tools. It’s struggling to keep one alive long enough for enterprise teams to finish onboarding onto it. If your organization is still running Q Developer, the April 30, 2027 deadline isn’t far off once you account for procurement, security review, and re-training cycles.
Google’s Gemini CLI Bait and Switch
Google’s move is subtler but just as disruptive. At I/O on May 19, 2026, Google announced it was transitioning Gemini CLI to a new agentic platform called Antigravity CLI, giving developers a 30-day migration window. That window closed June 18, 2026. After that date, Gemini CLI stopped working entirely for Google AI Pro, Ultra, and free-tier users.
The same cutoff hit Gemini Code Assist for GitHub: no new installations on GitHub organizations after June 18, and requests to existing installations stopped being served in the weeks that followed. The one group spared entirely is paying enterprise customers on a Gemini Code Assist Standard or Enterprise license. Their access carried on unchanged.
That split matters more than it looks. Free and individual-tier users, the developers with the least bargaining power, got pushed onto an unfamiliar platform with barely a month’s notice. FOSS Force summed up the reaction bluntly, running a piece titled “Gemini CLI’s Short Life and Google’s Antigravity Bait-and-Switch.” GitHub discussion threads show developers confused about losing paid subscriptions mid-cycle.
Antigravity itself, announced on the Google Developers Blog back in November 2025, isn’t a simple CLI update. It’s a structural break from the IDE-extension model Gemini Code Assist used, built around multiple AI agents that spawn, coordinate, and execute complex tasks autonomously. Some early adopters on user forums reported it “couldn’t do even simple stuff” as recently as January 2026, a reminder that agent-first rewrites don’t always ship stable on day one.
The $60 Billion Bombshell: SpaceX Buys Cursor
On June 16, 2026, four days after its own $75 billion IPO, SpaceX exercised an option to buy Anysphere, the parent company of Cursor, for $60 billion in an all-stock deal. It’s the largest venture-backed startup acquisition in history, filed with the SEC via Form 8-K, and it puts SpaceX in direct competition with Microsoft, Google, and Anthropic for developer mindshare.
The deal gives xAI, which merged with SpaceX in February 2026, its first serious entry into developer tools. It’s also a strange fit on paper: a rocket and satellite company now owns one of the most widely used AI coding assistants on the planet.
Cursor’s growth explains the price tag even if the buyer doesn’t. Annualized revenue went from roughly $100 million in 2024 to about $4 billion by June 2026, one of the fastest SaaS growth curves ever recorded, with roughly $2.6 billion of that coming from enterprise customers. But the deal closed while Cursor’s own market share was sliding. Corporate card spending data from Ramp shows Cursor’s share falling from about 41% in June 2025 to around 26% by May 2026, even as Anthropic’s Claude Code reportedly climbed toward 50% over the same stretch.
Two months before the acquisition closed, SpaceX had already moved Cursor’s compute onto xAI’s Colossus supercomputer, cutting its reliance on Anthropic and OpenAI models. That timing suggests this wasn’t a spontaneous bet. SpaceX was integrating Cursor before the ink was dry.
“What began as a race to deliver the most ‘magical’ developer experience is now evolving into a contest of operational excellence, commercial maturity, and enterprise readiness.”
Philip Walsh, Senior Director Analyst, Gartner
The deal isn’t finished yet. It’s expected to close in Q3 2026, subject to antitrust review, and the agreement carries a $10 billion termination fee alongside a separate $4 billion fee if it fails on antitrust grounds. Windsurf, a second independent AI-native IDE, was already absorbed by Cognition, maker of the Devin autonomous coding agent, earlier in 2026. Between the two deals, the independent AI-IDE category effectively disappeared within months.
Where GitHub Copilot Stands While Rivals Implode
Amid all this churn, GitHub Copilot just keeps growing. It’s crossed 20 million users, and GitHub’s own 2025 Octoverse report found over 1.1 million public repositories now import an LLM SDK, with 80% of new GitHub developers using Copilot within their first week. GitHub added more than 36 million developers in the 12 months to August 2025, its fastest growth rate ever, pushing total developers past 180 million, according to the GitHub Blog.
Market trackers put Copilot’s share of the AI coding tools category at roughly 37 to 42%, and GitHub says Copilot-enabled repositories now see about 46% of committed code generated with its help. That’s not a company defending territory. That’s a company that never stopped compounding while its rivals were busy discontinuing their own products.
The Numbers Behind the Chaos
Here’s the state of play across the four names that matter most right now.
Product
Parent
Status as of July 2026
Key figure
GitHub Copilot
Microsoft
Growing, market leader
20M+ users, ~37–42% share
Amazon Q Developer
AWS
Discontinued, replaced by Kiro
End of support April 30, 2027
Gemini CLI / Code Assist
Google
Free tier retired, enterprise protected
Cutover completed June 18, 2026
Cursor
SpaceX / xAI (pending)
Acquired for $60B, closing Q3 2026
~$4B ARR, share down to ~26%
Zoom out and the category itself is exploding even as individual products die. Gartner puts the enterprise AI coding agent market at $9.8 to $11 billion annualized as of April 2026, and the firm’s broader AI platforms and models forecast, published just three days before this article, projects worldwide spending will hit $64 billion in 2026, up 63.4% from $39 billion in 2025.
“Enterprise AI budgets are coming under greater scrutiny, with increased focus on usage efficiency, cost control and measurable outcomes.”
Arunasree Cheparthi, Senior Principal Research Analyst, Gartner
Worth flagging: market share figures for Copilot, Cursor, and Gemini vary by 5 to 10 points depending on whether the tracker is measuring revenue, seat count, or corporate spend data like Ramp’s. Treat any single number as a directional estimate, not a settled fact.
The Trust Problem Nobody’s Fixing
Bigger platforms and bigger checks don’t fix the thing developers actually complain about, which is that AI-generated code still isn’t reliable enough to trust blindly. Stack Overflow’s 2025 Developer Survey, nearly 49,000 respondents across 177 countries, found only 29 to 33% of developers trust the accuracy of AI-generated code, down from around 40 to 43% in 2024. Meanwhile, 84% of developers now use or plan to use AI tools regardless.
“One of the most notable trends in this year’s survey is the continued rise in AI tool usage, now at 84% of developers using or planning to use them, contrasted with a clear drop in favourability.”
Erin Yepis, Senior Analyst, Market Research & Insights, Stack Overflow
That gap between adoption and trust shows up in how the tools actually behave in production. GitClear’s analysis of 211 million lines of code found churn climbing from 3.1% to 5.7% while refactoring dropped from 25% of changes to under 10%. CodeRabbit separately found 2.74 times more security vulnerabilities in AI-coauthored pull requests. Speed is up. Code health, by these measures, is not keeping pace.
“One of the most common stories I hear in 2025 goes like this: someone gives an AI coding agent a try, expecting magic. But after a few actions, it messes up the architecture, changes something it shouldn’t, or just spits out bad code.”
Andrey Korchak, CTO, quoted in LeadDev
Our read: this signals the consolidation wave is really a bet on distribution and capital, not proof that any one platform has solved the reliability problem. Owning the market doesn’t mean the product got better. It means the company writing the checks has more time to figure it out.
There’s also a governance angle specific to the SpaceX deal. Existing unpatched vulnerabilities in Cursor now sit inside a company whose other assets include Starlink and satellite infrastructure, a supply chain concentration risk that traditional IDE acquisitions never really raised before.
What Engineering Leaders Should Do Now
If you’re a CTO, platform architect, or engineering lead making tooling decisions for 2026 and 2027, five things matter more than picking “the best” tool.
Treat forced migrations as a budget line, not a hypothetical. Any team still on Amazon Q Developer needs a Kiro migration plan on the books before April 30, 2027.
Understand what vendor lock-in means now. Choosing a coding assistant is increasingly a decision about whose training data pipeline your codebase feeds into, which is now a real data-governance question for regulated industries, not just a developer-experience preference.
Watch how IDE-optional the market gets. Gartner projects that more than 65% of engineering teams will treat IDEs as optional by 2027. That’s a forecast, not an observed trend yet, but it should shape whether you invest in IDE-first or spec-first workflows today.
Build verification discipline, not just tool preference. With trust in AI output near historic lows even as usage hits record highs, the real lever available to you is review process, not tool selection.
Read the enterprise contract terms closely. Google’s decision to shield paying Gemini Code Assist customers while cutting off free users during the Antigravity transition is a preview of how future shutdowns will likely be handled. Enterprise agreements are becoming the insulation layer against sudden product death.
For a deeper look at how the leading models stack up on raw coding performance, see our Claude Opus 4.8 vs GPT-5.6 coding model comparison, and for the capital side of this story, our breakdown of 2026 venture capital trends covers exactly the kind of funding trajectory that made Cursor an acquisition target in the first place.
Frequently Asked Questions
Is Amazon Q Developer being discontinued?
Yes. AWS blocked new Amazon Q Developer signups on May 15, 2026, and will end support for its IDE plugins and paid subscriptions on April 30, 2027. AWS is migrating users to Kiro, a new spec-driven agentic IDE built on Code OSS.
What replaced Gemini CLI?
Google replaced Gemini CLI with Antigravity CLI starting June 18, 2026, after a 30-day migration window announced at I/O. Enterprise users on Gemini Code Assist Standard or Enterprise licenses keep unchanged access, while free and individual-tier users must migrate to Antigravity.
Who bought Cursor?
SpaceX acquired Anysphere, the parent company of Cursor, for $60 billion in an all-stock deal announced June 16, 2026. It’s the largest venture-backed startup acquisition on record, expected to close in Q3 2026 pending regulatory approval.
What percentage of the market does GitHub Copilot have?
Estimates place GitHub Copilot’s share of the AI coding tools market at roughly 37 to 42% as of 2026, based on multiple market-tracking reports, though figures vary depending on whether share is measured by revenue, seats, or corporate spend data.
Do developers trust AI-generated code?
Not really. Stack Overflow’s 2025 Developer Survey of nearly 49,000 developers found only 29 to 33% trust AI output accuracy, down from about 40 to 43% in 2024, even as tool usage climbed to 84%, a widening gap between adoption and trust.
What to Watch Next
Here’s what you now understand that you probably didn’t ten minutes ago: the “Big Tech owns AI coding tools” story isn’t really about Big Tech tightening a grip it already had. It’s about two products dying inside their own companies and one $60 billion acquisition that hands a rocket company control of the market’s most talked-about coding assistant.
Over the next 6 to 18 months, watch three things specifically. First, whether the SpaceX-Cursor deal clears antitrust review in Q3 2026, and what conditions regulators attach if it does. Second, whether Kiro’s early 250,000-user traction holds up once Amazon Q Developer’s paying customers are forced to migrate rather than choosing to. Third, whether Antigravity’s early stability complaints fade or harden into a genuine reputation problem for Google’s agent-first bet.
None of that tells you which tool to pick. It tells you that the tool you pick today probably won’t be the tool your team is using in 2027, no matter who makes it.
Best LLM Evaluation Tools 2026: 7 Tested, Ranked, Compared
ML Tooling / Developer Focus
Best LLM Evaluation Tools 2026: 7 Tested, Compared
By the NeuralWired ML Team · Updated July 23, 2026 · Part of our ongoing Tested series
Your RAG pipeline passed every demo. Then it hit production and started citing the wrong policy document to real customers, and nobody noticed for six days. That gap between “looked fine in the demo” and “actually correct at scale” is exactly why best LLM evaluation tools 2026 has become one of the most searched phrases among ML teams this year. We tested seven of them against real traces, real budgets, and real failure modes, so you don’t have to guess which one fits your stack.
Three things collided in 2026 to push evaluation tooling from “nice to have” to “line item in the budget.” First, the money: Gartner’s own newsroom forecasts the global GenAI models market will top 25 billion dollars in 2026 and reach 75 billion by 2029, with LLM observability investment climbing to half of all GenAI deployments by 2028, up from roughly 15 percent in early 2026. Second, the regulation: EU AI Act enforcement begins in August 2026, and several compliance analysts now cite documented evaluation practice as a direct requirement for systems serving EU users, not a suggestion.
Third, and this is the one engineers actually feel: the classic benchmarks stopped telling you anything. MMLU and HellaSwag scores for frontier models now cluster above 88 percent, close enough that the differences sit inside measurement noise. If you’re still leaning on those numbers to pick a model or judge, you’re comparing rounding errors. That’s part of why harder, contamination-resistant suites like GPQA-Diamond, SWE-bench Verified, and LiveBench have taken over as the benchmarks that actually separate models.
The number to budget against: running DeepEval and RAGAS together across 10,000 RAG traces a day costs roughly 200 to 600 dollars a month in LLM-judge token spend on GPT-4o, according to a cost-modeling analysis from genai.qa. Swap in a cheaper judge like GPT-4o-mini or Claude Haiku and that drops 60 to 80 percent.
And here’s the part most “best tools” roundups skip: plenty of teams still haven’t automated this at all. A LangChain survey of more than 1,300 practitioners found 59.8 percent still rely on human review to grade outputs, while 53.3 percent use LLM-as-a-judge, with heavy overlap between the two. The share of teams doing no testing whatsoever did drop, from 29.5 percent to 22.8 percent, but that’s still nearly a quarter of production teams flying blind.
The 7 tools, ranked and compared
We grouped these by what they’re actually built for, because “best overall” is the wrong question. A code-first library that lives in your CI pipeline solves a different problem than a hosted platform your product team logs into. Picking between them is now an architecture decision made at design time, not a bolt-on before launch.
Tool
Best for
Model
Standout feature
DeepEval
CI/CD test suites, pytest-style workflows
Open source
50+ metrics, built by Confident AI’s Jeffrey Ip and Kritin Vongthongsri
RAGAS
RAG-specific scoring
Open source
Reference-free faithfulness, relevancy, and context metrics; academic roots
Promptfoo
Red-teaming, prompt regression
Open source
YAML config, 40+ adversarial testing plugins
Langfuse
Production tracing, self-hosted teams
Open source (MIT), acquired by ClickHouse
Free self-hosted tier, no usage cap
Arize Phoenix
OpenTelemetry-native tracing
Open source + commercial (Arize AX)
Built by former Uber Michelangelo lead Aparna Dhinakaran’s team
Braintrust
Full eval-to-production CI platform
Commercial
Free tier: 1M spans/month, 10K evals
Galileo AI
Regulated, high-volume production
Commercial
Luna-2 judge models for sub-200ms scoring at scale
DeepEval vs RAGAS: the comparison everyone actually searches for
RAGAS is the specialist. It was built for one job, scoring retrieval-augmented generation, and it does that job with four metrics that don’t require a fixed reference answer: faithfulness, answer relevancy, context precision, and context recall. It traces back to a 2023 research paper that reportedly got a mention from OpenAI at a developer event that year, which is part of why it still carries more academic weight than most commercial entrants launched since.
DeepEval is the generalist. It handles RAG too, but also agents, chatbots, and general-purpose CI/CD test gates, all in a pytest-style workflow your existing test suite already understands. If you’re only doing RAG, RAGAS is narrower and arguably sharper. If you’re shipping agents and chatbots alongside RAG, DeepEval covers more ground without forcing you into three separate frameworks. Most teams we found running mature pipelines use both, RAGAS for the retrieval layer, DeepEval for everything wrapped around it.
The observability layer: Langfuse vs Arize Phoenix
Langfuse got acquired by ClickHouse in January 2026, and the open-source repo has kept shipping since, with the self-hosted, MIT-licensed version still free and uncapped. Phoenix, Arize’s open-source tracing library, is OpenTelemetry-native, which matters if your infrastructure team already standardized on OTel for everything else. One licensing note worth flagging honestly: sources describe Phoenix’s license inconsistently, some say Apache 2.0, others Elastic License 2.0, so confirm the current terms directly against the Arize-ai/phoenix repository before you build a dependency on it.
How accurate is LLM-as-a-judge, really
Nearly every tool on this list leans on the same underlying mechanism: one model scoring another’s output against a rubric instead of a fixed correct answer. Aggregated studies put that agreement with human raters somewhere between 80 and 92 percent, at a fraction of the cost of full manual review, roughly 500 to 5,000 times cheaper depending on the study.
That’s genuinely good. It is not perfect, and treating it as a solved problem is where teams get burned. Position bias, verbosity bias (judges tend to prefer longer answers even when they’re not better), and self-preference bias, where a judge model rates outputs from its own model family more favorably, are all documented and unresolved.
“Marketing, not science.”
Nathan Lambert, AI researcher, Interconnects.ai, on cross-vendor benchmark comparisons that can’t be independently verified. Read the original piece
Lambert’s line is from December 2023, and it’s still getting cited in curated 2026 evaluation reading lists for a reason. His point wasn’t about the tools on this list specifically, it was about competitors publishing benchmark claims about each other’s models without access to verify them. The same skepticism applies here: DeepEval, Braintrust, and Latitude all publish comparison content that ranks their own product first. Worth remembering while you read anyone’s “best tools” list, including this one.
What Husain and Dhinakaran are actually saying
Hamel Husain, the independent ML consultant co-authoring the upcoming O’Reilly book on AI evaluation with Shreya Shankar, has argued in multiple interviews that teams who appear to ship without formal evals are usually leaning on evaluation work someone else already did upstream, most often the model provider’s own internal testing. His and Shankar’s broader position, laid out across their public evals masterclass materials, warns specifically against fully automating evaluation without keeping a human grounded in product-specific context.
Aparna Dhinakaran, Arize’s co-founder and chief product officer, framed the shift differently at Arize’s Observe 2026 event: the industry is moving away from a person manually reading individual traces one by one, toward a person overseeing a fleet of agents that check each other’s work. It’s a more optimistic read than Husain’s, and both are right about different parts of the pipeline.
The stack recipes teams actually run
No single tool here covers RAG, agents, and chatbots equally well. The category is fragmented on purpose, and stacking two or three tools for different failure modes is standard practice among teams that have been doing this for a while, not a sign that someone picked wrong the first time. The pattern we saw repeated most often:
CI gate: DeepEval or Promptfoo, catching regressions before merge
RAG-specific scoring: RAGAS, layered in wherever retrieval is involved
Production sampling: Langfuse or Phoenix, watching what real traffic actually does
That combination gets most of a full commercial platform’s coverage for a fraction of the price. If your team is already dealing with agent deployments that keep failing in production, this is the stack worth building before you shop for anything commercial.
Where the hype outruns the evidence
A few things this space glosses over that are worth saying plainly. First, benchmark saturation is real, not a talking point. Second, an 80 to 92 percent agreement rate with human raters means a judge model is wrong on a meaningful slice of calls, and none of the vendor pages selling “automated evaluation at scale” put that number front and center. Third, the timeline claim that any single tool “solves” LLM evaluation is overstated across nearly every vendor’s own marketing page, including the ones cited in this article.
Our read: the risk scenario worth planning around isn’t picking the wrong tool. It’s picking a commercial platform, trusting its automated pass rate completely, and losing the human error-analysis step that would have caught something product-specific a rubric never would.
FAQ
What is the best LLM evaluation tool in 2026?
There’s no single best tool, it depends on use case. DeepEval and Promptfoo lead for CI/CD-integrated testing, RAGAS is the standard for RAG-specific metrics, and Langfuse or Arize Phoenix lead production tracing. Most mature teams combine two or three tools rather than relying on one platform.
Is RAGAS or DeepEval better for RAG evaluation?
RAGAS is purpose-built for RAG with four research-backed reference-free metrics: faithfulness, answer relevancy, context precision, and context recall. DeepEval covers RAG plus agents, chatbots, and CI/CD integration, making it broader but less specialized. Many teams pair both.
How much does LLM evaluation cost at scale?
Running DeepEval and RAGAS together against 10,000 RAG traces a day costs roughly 200 to 600 dollars a month in LLM-judge API spend on GPT-4o. Switching to cheaper judge models like GPT-4o-mini or Claude Haiku cuts that by 60 to 80 percent, with a manageable accuracy trade-off.
What is LLM-as-a-judge and how accurate is it?
LLM-as-a-judge uses one LLM to score another’s output against a rubric instead of comparing it to a fixed answer. Published studies put agreement with human raters at roughly 80 to 92 percent, at a fraction of the cost of full human review, but it carries known biases like favoring longer or first-listed answers.
Are MMLU and other classic benchmarks still useful in 2026?
Only marginally for frontier models. MMLU and HellaSwag scores have saturated above 88 percent, with differences between top models falling inside measurement noise. Harder, contamination-resistant benchmarks like GPQA-Diamond, SWE-bench Verified, and LiveBench are now more informative for comparing leading LLMs.
Is Langfuse free?
Langfuse’s self-hosted version is free and MIT-licensed with no usage limits. The managed cloud version has a free Hobby tier of 50,000 units a month, then paid tiers starting around 29 to 199 dollars a month depending on retention and support needs.
Where this goes next
What you now know that most “top 7 tools” listicles won’t tell you: there is no finish line here. The tools on this list are converging, ClickHouse now owns Langfuse, commercial platforms keep absorbing open-source primitives, and the EU AI Act’s August enforcement date is going to force teams that have been skipping documented evaluation to catch up fast.
Three things worth watching over the next 6 to 18 months:
Whether Galileo and Braintrust’s judge-model approach (cheaper, faster, purpose-built judges instead of general-purpose GPT-4o calls) becomes the default rather than the exception
Whether agent-specific, multi-turn evaluation standards mature enough to replace the single-turn accuracy checks most of this tooling was originally built around
Whether EU AI Act enforcement actually changes tool selection, or just adds a compliance checkbox on top of whatever teams were already running
If you’re picking a stack this quarter: start with the CI gate, add the RAG-specific layer only if you’re running RAG, and don’t buy a commercial platform until you’ve felt the limits of the free, open-source combination first.
Nearly 3 in 10 Small Businesses Hit by Deepfake Scams in 2026
NeuralWired Cybersecurity Desk · Published July 23, 2026
In February 2024, a finance employee at UK engineering firm Arup joined what looked like a routine video call with the CFO and several colleagues. He wired $25.6 million across 15 transactions before anyone realized every face on that call except his own was AI generated. Two years later, that trick has trickled all the way down to businesses with a dozen employees and no IT department: 29% of small businesses now say they’ve experienced a deepfake scam in the past year, according to a new survey from cybersecurity firm VikingCloud.
That number, buried inside VikingCloud’s 2026 SMB Threat Landscape Report, is the clearest signal yet that deepfake fraud stopped being an enterprise problem sometime in the last eighteen months. It’s now a Tuesday-afternoon problem for a plumbing company in Ohio or a marketing agency in Manchester. And the FBI, for the first time in its Internet Crime Complaint Center’s roughly 25-year history, agrees the threat is big enough to track on its own.
VikingCloud surveyed small business owners and operators for its 2026 threat report, and the results reorder what SMBs are worried about. More than a quarter said they’d experienced a deepfake scheme (29%), a customer data breach (27%), a ransomware attack (26%), or a denial of service attack (26%) in the past year. Taken together, 75% of SMB owners now rank cyberattacks as their number one operational threat for 2026, the first time in this survey series that cybersecurity has outranked economic pressure. Forty percent said a cyberattack costing $100,000 or less could put them out of business entirely.
A note on the source VikingCloud hasn’t published full survey methodology, sample size, or margin of error in its public summary; the underlying data sits behind a lead-gen form. That doesn’t make the 29% figure false, but it means it should be read as “according to a vendor survey of small business owners,” not as census-grade data. Compare it against the FBI figure below, which is independently audited.
The FBI Just Made It Official
For 2025, the FBI’s Internet Crime Complaint Center broke out AI-enabled fraud as its own standalone category for the first time. IC3 logged 22,364 complaints with a reported AI nexus, totaling $893,346,472 in adjusted losses, according to the FBI IC3 2025 Annual Report published in April 2026. That figure is the closest thing this space has to a government-audited number, and it’s worth breaking down by category.
Fraud category (AI referenced)
2025 adjusted losses
Investment fraud
$632.0 million
Business email compromise
$30.3 million
Tech and customer-support scams
$19.5 million
Confidence and romance scams
$19.0 million
Employment scams
$12.6 million
Business email compromise is the line that should matter most to a small business owner. It’s the category built entirely around impersonating someone the victim already trusts, a vendor, a boss, a bank contact, and it’s exactly the mechanism behind the Arup case.
The Case That Changed Everything: Arup’s $25.6 Million Call
Arup’s Hong Kong finance team received what appeared to be a standard request from the company’s UK-based CFO: move funds for a confidential transaction. The employee had doubts, so he did what security training tells you to do. He joined a video call to verify. Every other participant on that call, including the person who looked and sounded like the CFO, was an AI-generated deepfake. He made the transfers. Reporting from the Financial Times and CNN in May 2024 confirmed the total loss at $25.6 million across 15 wire transactions, and the case has become the reference point every security vendor cites when explaining why video verification alone is no longer enough.
Arup is a global engineering firm with sophisticated finance operations, not a small business. That distinction matters, and we’ll come back to it. But the mechanics of the attack, real-time video and voice synthesis convincing enough to fool someone who was actively trying to verify, work exactly the same way against a five-person accounting team as they did against Arup’s.
When the Defense Works: WPP’s Near Miss
Not every attempt succeeds, and the counter-example is worth knowing. Scammers targeted WPP CEO Mark Read using a cloned voice and a spoofed Microsoft Teams meeting invite, built around a fake WhatsApp account using his public photo, according to an entry in the OECD.AI Incident Database and reporting from Marketing-Interactive. Staff escalated before any money moved. WPP confirmed zero losses.
What stopped it wasn’t detection software. It was a human asking a question the scammer couldn’t answer and refusing to proceed until someone verified through a separate channel. That’s a cheap lesson, and it’s the same one at the center of the advice section below.
Why Small Businesses Are the Easier Target
Here’s the uncomfortable part for small business owners: being small isn’t protection. It’s the opposite. VikingCloud’s data shows 84% of SMB owners self-manage their own cybersecurity, with no dedicated IT or security staff. That means the same person approving a vendor invoice is also the last line of defense against a fraudulent one, with no gatekeeper, no second sign-off, no layered approval chain to slow things down.
An enterprise like Arup still has structural weaknesses attackers can exploit, but it also has finance controls, compliance teams, and escalation paths. A twelve-person business usually has one bookkeeper and a Slack channel. Attackers know which door is easier to walk through.
“We only have like one really good example in the news right now of that organization in Hong Kong that ended up falling for and sending $25 million based on a deepfake audio and video scam, and I think we’re going to see a lot more business email compromise style events because of AI.”Rachel Tobac, CEO, SocialProof Security · 8th Layer Insights podcast, The Cyber Wire, April 9, 2024
Tobac’s prediction has aged into the current data. The FBI’s BEC-with-AI-nexus figure alone hit $30.3 million in 2025, and that’s before counting the cases that never get formally reported, which fraud researchers generally assume is the majority of them.
“AI-generated media is not just a future risk, it’s a real business threat. We’re seeing executives impersonated, hiring processes compromised, and financial safeguards bypassed with alarming ease.”Tony Lee, Head of Consulting, Hong Kong & Macau, Trend Micro · Media OutReach Newswire, July 10, 2025
Worth flagging: Lee’s employer, Trend Micro, sells deepfake detection tools, so treat the quote as an informed but interested voice rather than a neutral one.
Can You Trust Your Own Eyes?
Most SMB owners assume they’d notice if something felt off on a call. The data says otherwise. Controlled lab studies compiled by security research firm DeepStrike found human accuracy at spotting high-quality deepfake video sits at just 24.5%, even though roughly 60% of people believe they could identify one. That gap between confidence and competence is arguably the more dangerous number in this whole story.
The technical barrier to producing convincing fakes keeps dropping too. McAfee’s consumer research found a voice clone with about 85% similarity to the original can now be generated from just three seconds of audio, easily pulled from a podcast clip, a local news interview, or a company’s own marketing video.
The Regulatory Clock Is Ticking
Two regulatory shifts land right around this article’s publish date. The EU AI Act’s Article 50 transparency rules, requiring disclosure and labeling of AI-generated content, take effect in August 2026, with penalties reaching €35 million or 7% of global turnover for noncompliance. Meanwhile, roughly 46 to 47 US states have now passed some form of deepfake-specific legislation, spanning election-related disclosure rules, non-consensual imagery protections, and fraud statutes, according to MultiState’s legislative tracking.
None of this stops a scam call from reaching a small business tomorrow morning. But it does signal that lawmakers on both sides of the Atlantic have stopped treating deepfakes as a novelty problem.
Reader Beware: Not Every Stat Holds Up
Scroll through enough 2026 deepfake coverage and you’ll hit percentage increases that sound apocalyptic: 2,137%, 3,892%, four-digit growth claims stacked one after another. A research team at Digital Applied spent its July 2026 audit picking these apart, arguing that the field is crowded with numbers nobody actually verifies, loss figures with no traceable primary source, surge percentages that contradict each other depending on which vendor published them, and forecasts that get recycled as if they were measurements.
Our read: most of those huge percentage jumps are real in direction but misleading in scale. A fraud category that goes from 0.1% to 6.5% of total fraud attempts, which is roughly what’s happened according to fraud-detection firm Signicat, produces an enormous percentage increase almost automatically, simply because it started near zero. That’s still a genuine and fast-growing threat. It’s just not the same thing as the flat “up 3,892% this year” headline that gets repeated without context.
It’s also worth being honest about scale. Most of the largest documented deepfake losses, Arup’s $25.6 million among them, hit large enterprises with the kind of finance operations that can move eight figures in a single transfer. A small business physically can’t lose that much in one incident. The realistic SMB exposure looks more like tens of thousands of dollars per event, which is still enough to close a business operating on thin margins, but the “small businesses are next in line for a $25 million loss” framing overstates the individual stakes even while understating how often SMBs get hit.
The One Habit That Beats the Software
Security researchers keep landing on the same conclusion, and it isn’t a product pitch. Verizon’s Data Breach Investigations Report, cited across multiple 2026 industry analyses, consistently finds the human element involved in more than 60% of breaches. A basic callback-verification habit defeats a deepfake exactly as well as it defeats a decades-old phone scam, because the fake voice or face is only dangerous if the person on the other end skips the second check.
Set a callback rule. Any request to move money, change banking details, or reset credentials gets verified by calling a number pulled from your own records, never one supplied in the suspicious message or call.
Agree on a code word. A pre-shared phrase for high-stakes requests costs nothing and a real-time deepfake can’t guess it.
Slow down on urgency. Scammers manufacture time pressure because it stops people from verifying. Treat “this has to happen right now” as the red flag it is.
Train the one person who approves payments. If your business doesn’t have a finance team, whoever signs off on transfers is your entire defense layer. Make sure they know this playbook exists.
Gartner had already predicted where this was heading: by 2026, the firm projected that 40% of enterprises would stop trusting standalone identity verification because of deepfakes. That prediction is landing now, and the fix it points to isn’t more software, it’s a second channel that a synthetic voice or face can’t fake its way through.
Frequently Asked Questions
What percentage of small businesses have experienced a deepfake scam?
According to VikingCloud’s 2026 SMB Threat Landscape Report, 29% of small businesses reported experiencing a deepfake scheme in the past 12 months, making it one of the most common cyber incidents SMB owners now report, alongside data breaches and ransomware.
How much money has been lost to deepfake and AI-enabled fraud in 2025?
The FBI’s Internet Crime Complaint Center logged $893,346,472 in adjusted losses from 22,364 US complaints referencing AI in 2025, the first year the FBI tracked AI-enabled fraud as its own standalone category.
How can a small business protect itself from deepfake scams?
Require a second-channel verification, a callback to an internally stored phone number or a pre-agreed code word, for any request involving wire transfers, banking-detail changes, or credential resets, even ones that arrive by video call. It consistently ranks above detection software as the lowest-cost, most effective defense.
Why are small businesses targeted by deepfake scammers more than large companies?
Small businesses often rely on informal, trust-based approval processes with no dedicated IT or security staff. Eighty-four percent of SMB owners self-manage their own cybersecurity, per VikingCloud’s 2026 report, which removes the layered sign-off chain that would otherwise catch a fraudulent request.
Can humans reliably spot a deepfake video?
No. Controlled studies find human accuracy at identifying high-quality deepfake videos is only about 24.5%, even though roughly 60% of people believe they could spot one, a gap that itself increases risk by creating false confidence.
Where This Goes Next
Two things are converging right now that weren’t true even a year ago. The FBI has an audited number to point to for the first time, and small business owners are, for the first time in this survey series, ranking cyberattacks above the economy as their biggest worry. Neither of those happens without the other. Watch three things over the next six to eighteen months: whether EU AI Act enforcement actually produces fines large enough to change vendor behavior, whether cyber insurers start pricing deepfake-specific BEC into small business premiums, and whether the “29%” figure gets replicated by a source willing to publish full methodology.
The takeaway for anyone running a small business isn’t to panic about AI. It’s to put a five-minute verification habit in place before you need it. The businesses in the Arup and WPP stories both had smart people on the call. Only one of them had a process that didn’t depend on trusting what they saw.
Nearly 3 in 10 Small Businesses Hit by Deepfake Scams in 2026
NeuralWired Cybersecurity Desk · Published July 23, 2026
In February 2024, a finance employee at UK engineering firm Arup joined what looked like a routine video call with the CFO and several colleagues. He wired $25.6 million across 15 transactions before anyone realized every face on that call except his own was AI generated. Two years later, that trick has trickled all the way down to businesses with a dozen employees and no IT department: 29% of small businesses now say they’ve experienced a deepfake scam in the past year, according to a new survey from cybersecurity firm VikingCloud.
That number, buried inside VikingCloud’s 2026 SMB Threat Landscape Report, is the clearest signal yet that deepfake fraud stopped being an enterprise problem sometime in the last eighteen months. It’s now a Tuesday-afternoon problem for a plumbing company in Ohio or a marketing agency in Manchester. And the FBI, for the first time in its Internet Crime Complaint Center’s roughly 25-year history, agrees the threat is big enough to track on its own.
VikingCloud surveyed small business owners and operators for its 2026 threat report, and the results reorder what SMBs are worried about. More than a quarter said they’d experienced a deepfake scheme (29%), a customer data breach (27%), a ransomware attack (26%), or a denial of service attack (26%) in the past year. Taken together, 75% of SMB owners now rank cyberattacks as their number one operational threat for 2026, the first time in this survey series that cybersecurity has outranked economic pressure. Forty percent said a cyberattack costing $100,000 or less could put them out of business entirely.
A note on the source
VikingCloud hasn’t published full survey methodology, sample size, or margin of error in its public summary; the underlying data sits behind a lead-gen form. That doesn’t make the 29% figure false, but it means it should be read as “according to a vendor survey of small business owners,” not as census-grade data. Compare it against the FBI figure below, which is independently audited.
The FBI Just Made It Official
For 2025, the FBI’s Internet Crime Complaint Center broke out AI-enabled fraud as its own standalone category for the first time. IC3 logged 22,364 complaints with a reported AI nexus, totaling $893,346,472 in adjusted losses, according to the FBI IC3 2025 Annual Report published in April 2026. That figure is the closest thing this space has to a government-audited number, and it’s worth breaking down by category.
Fraud category (AI referenced)
2025 adjusted losses
Investment fraud
$632.0 million
Business email compromise
$30.3 million
Tech and customer-support scams
$19.5 million
Confidence and romance scams
$19.0 million
Employment scams
$12.6 million
Business email compromise is the line that should matter most to a small business owner. It’s the category built entirely around impersonating someone the victim already trusts, a vendor, a boss, a bank contact, and it’s exactly the mechanism behind the Arup case.
The Case That Changed Everything: Arup’s $25.6 Million Call
Arup’s Hong Kong finance team received what appeared to be a standard request from the company’s UK-based CFO: move funds for a confidential transaction. The employee had doubts, so he did what security training tells you to do. He joined a video call to verify. Every other participant on that call, including the person who looked and sounded like the CFO, was an AI-generated deepfake. He made the transfers. Reporting from the Financial Times and CNN in May 2024 confirmed the total loss at $25.6 million across 15 wire transactions, and the case has become the reference point every security vendor cites when explaining why video verification alone is no longer enough.
Arup is a global engineering firm with sophisticated finance operations, not a small business. That distinction matters, and we’ll come back to it. But the mechanics of the attack, real-time video and voice synthesis convincing enough to fool someone who was actively trying to verify, work exactly the same way against a five-person accounting team as they did against Arup’s.
When the Defense Works: WPP’s Near Miss
Not every attempt succeeds, and the counter-example is worth knowing. Scammers targeted WPP CEO Mark Read using a cloned voice and a spoofed Microsoft Teams meeting invite, built around a fake WhatsApp account using his public photo, according to an entry in the OECD.AI Incident Database and reporting from Marketing-Interactive. Staff escalated before any money moved. WPP confirmed zero losses.
What stopped it wasn’t detection software. It was a human asking a question the scammer couldn’t answer and refusing to proceed until someone verified through a separate channel. That’s a cheap lesson, and it’s the same one at the center of the advice section below.
Why Small Businesses Are the Easier Target
Here’s the uncomfortable part for small business owners: being small isn’t protection. It’s the opposite. VikingCloud’s data shows 84% of SMB owners self-manage their own cybersecurity, with no dedicated IT or security staff. That means the same person approving a vendor invoice is also the last line of defense against a fraudulent one, with no gatekeeper, no second sign-off, no layered approval chain to slow things down.
An enterprise like Arup still has structural weaknesses attackers can exploit, but it also has finance controls, compliance teams, and escalation paths. A twelve-person business usually has one bookkeeper and a Slack channel. Attackers know which door is easier to walk through.
“We only have like one really good example in the news right now of that organization in Hong Kong that ended up falling for and sending $25 million based on a deepfake audio and video scam, and I think we’re going to see a lot more business email compromise style events because of AI.”
Rachel Tobac, CEO, SocialProof Security · 8th Layer Insights podcast, The Cyber Wire, April 9, 2024
Tobac’s prediction has aged into the current data. The FBI’s BEC-with-AI-nexus figure alone hit $30.3 million in 2025, and that’s before counting the cases that never get formally reported, which fraud researchers generally assume is the majority of them.
“AI-generated media is not just a future risk, it’s a real business threat. We’re seeing executives impersonated, hiring processes compromised, and financial safeguards bypassed with alarming ease.”
Tony Lee, Head of Consulting, Hong Kong & Macau, Trend Micro · Media OutReach Newswire, July 10, 2025
Worth flagging: Lee’s employer, Trend Micro, sells deepfake detection tools, so treat the quote as an informed but interested voice rather than a neutral one.
Can You Trust Your Own Eyes?
Most SMB owners assume they’d notice if something felt off on a call. The data says otherwise. Controlled lab studies compiled by security research firm DeepStrike found human accuracy at spotting high-quality deepfake video sits at just 24.5%, even though roughly 60% of people believe they could identify one. That gap between confidence and competence is arguably the more dangerous number in this whole story.
The technical barrier to producing convincing fakes keeps dropping too. McAfee’s consumer research found a voice clone with about 85% similarity to the original can now be generated from just three seconds of audio, easily pulled from a podcast clip, a local news interview, or a company’s own marketing video.
The Regulatory Clock Is Ticking
Two regulatory shifts land right around this article’s publish date. The EU AI Act’s Article 50 transparency rules, requiring disclosure and labeling of AI-generated content, take effect in August 2026, with penalties reaching €35 million or 7% of global turnover for noncompliance. Meanwhile, roughly 46 to 47 US states have now passed some form of deepfake-specific legislation, spanning election-related disclosure rules, non-consensual imagery protections, and fraud statutes, according to MultiState’s legislative tracking.
None of this stops a scam call from reaching a small business tomorrow morning. But it does signal that lawmakers on both sides of the Atlantic have stopped treating deepfakes as a novelty problem.
Reader Beware: Not Every Stat Holds Up
Scroll through enough 2026 deepfake coverage and you’ll hit percentage increases that sound apocalyptic: 2,137%, 3,892%, four-digit growth claims stacked one after another. A research team at Digital Applied spent its July 2026 audit picking these apart, arguing that the field is crowded with numbers nobody actually verifies, loss figures with no traceable primary source, surge percentages that contradict each other depending on which vendor published them, and forecasts that get recycled as if they were measurements.
Our read: most of those huge percentage jumps are real in direction but misleading in scale. A fraud category that goes from 0.1% to 6.5% of total fraud attempts, which is roughly what’s happened according to fraud-detection firm Signicat, produces an enormous percentage increase almost automatically, simply because it started near zero. That’s still a genuine and fast-growing threat. It’s just not the same thing as the flat “up 3,892% this year” headline that gets repeated without context.
It’s also worth being honest about scale. Most of the largest documented deepfake losses, Arup’s $25.6 million among them, hit large enterprises with the kind of finance operations that can move eight figures in a single transfer. A small business physically can’t lose that much in one incident. The realistic SMB exposure looks more like tens of thousands of dollars per event, which is still enough to close a business operating on thin margins, but the “small businesses are next in line for a $25 million loss” framing overstates the individual stakes even while understating how often SMBs get hit.
The One Habit That Beats the Software
Security researchers keep landing on the same conclusion, and it isn’t a product pitch. Verizon’s Data Breach Investigations Report, cited across multiple 2026 industry analyses, consistently finds the human element involved in more than 60% of breaches. A basic callback-verification habit defeats a deepfake exactly as well as it defeats a decades-old phone scam, because the fake voice or face is only dangerous if the person on the other end skips the second check.
Set a callback rule. Any request to move money, change banking details, or reset credentials gets verified by calling a number pulled from your own records, never one supplied in the suspicious message or call.
Agree on a code word. A pre-shared phrase for high-stakes requests costs nothing and a real-time deepfake can’t guess it.
Slow down on urgency. Scammers manufacture time pressure because it stops people from verifying. Treat “this has to happen right now” as the red flag it is.
Train the one person who approves payments. If your business doesn’t have a finance team, whoever signs off on transfers is your entire defense layer. Make sure they know this playbook exists.
Gartner had already predicted where this was heading: by 2026, the firm projected that 40% of enterprises would stop trusting standalone identity verification because of deepfakes. That prediction is landing now, and the fix it points to isn’t more software, it’s a second channel that a synthetic voice or face can’t fake its way through.
Frequently Asked Questions
What percentage of small businesses have experienced a deepfake scam?
According to VikingCloud’s 2026 SMB Threat Landscape Report, 29% of small businesses reported experiencing a deepfake scheme in the past 12 months, making it one of the most common cyber incidents SMB owners now report, alongside data breaches and ransomware.
How much money has been lost to deepfake and AI-enabled fraud in 2025?
The FBI’s Internet Crime Complaint Center logged $893,346,472 in adjusted losses from 22,364 US complaints referencing AI in 2025, the first year the FBI tracked AI-enabled fraud as its own standalone category.
How can a small business protect itself from deepfake scams?
Require a second-channel verification, a callback to an internally stored phone number or a pre-agreed code word, for any request involving wire transfers, banking-detail changes, or credential resets, even ones that arrive by video call. It consistently ranks above detection software as the lowest-cost, most effective defense.
Why are small businesses targeted by deepfake scammers more than large companies?
Small businesses often rely on informal, trust-based approval processes with no dedicated IT or security staff. Eighty-four percent of SMB owners self-manage their own cybersecurity, per VikingCloud’s 2026 report, which removes the layered sign-off chain that would otherwise catch a fraudulent request.
Can humans reliably spot a deepfake video?
No. Controlled studies find human accuracy at identifying high-quality deepfake videos is only about 24.5%, even though roughly 60% of people believe they could spot one, a gap that itself increases risk by creating false confidence.
Where This Goes Next
Two things are converging right now that weren’t true even a year ago. The FBI has an audited number to point to for the first time, and small business owners are, for the first time in this survey series, ranking cyberattacks above the economy as their biggest worry. Neither of those happens without the other. Watch three things over the next six to eighteen months: whether EU AI Act enforcement actually produces fines large enough to change vendor behavior, whether cyber insurers start pricing deepfake-specific BEC into small business premiums, and whether the “29%” figure gets replicated by a source willing to publish full methodology.
The takeaway for anyone running a small business isn’t to panic about AI. It’s to put a five-minute verification habit in place before you need it. The businesses in the Arup and WPP stories both had smart people on the call. Only one of them had a process that didn’t depend on trusting what they saw.
Big Tech Layoffs 2026: Why AI Capex Explains It All
Big Tech
Big Tech Layoffs 2026: Why AI Capex Explains It All
Microsoft posted its best quarter ever and cut 4,800 jobs in the same three months. Amazon hit a record 13.1% operating margin and eliminated 30,000 corporate roles. Cisco broke its own revenue record and announced 4,000 layoffs the same week. None of that is a coincidence, and none of it is really about saving money on payroll either. It’s about where the money is actually going.
Big tech layoffs in 2026 keep landing next to record earnings, and the pattern only makes sense once you put the two numbers side by side: what these companies are cutting from headcount, and what they’re pouring into AI infrastructure. The gap between those numbers is the story.
Start with Meta, because the comparison is cleanest there and it sets the pattern for everyone else. Meta’s 2026 capital expenditure guidance sits at $125 billion to $145 billion. Its entire human compensation bill, salaries, benefits, equity, all of it, runs around $27 billion. Even if Meta fired every single employee tomorrow, it wouldn’t cover a fifth of what it’s already committed to spend on AI infrastructure. That comparison comes from a Yahoo Finance analysis of company disclosures, and it’s the single most useful number in this entire story.
Apply the same logic to Microsoft, Amazon, and Cisco and the picture holds. These aren’t companies trimming staff to fund a data center. They’re companies redirecting capital toward compute at a scale where headcount decisions barely register on the balance sheet.
Why this matters: If layoffs were really about cost savings, the numbers would be close. They’re not. Payroll cuts save these companies low single-digit billions. AI capex commitments run into the hundreds of billions. Two completely different orders of magnitude, decided by two largely separate processes inside the same company.
Company
2026 AI Capex
Jobs Cut
Same-Quarter Result
Microsoft
~$190B (guided)
4,800 (plus 9,100 prior round)
Record $82.89B quarterly revenue
Amazon
~$200B (guided)
~30,000 corporate roles
Record 13.1% operating margin
Cisco
$9B AI orders (raised guidance)
Fewer than 4,000 (~5%)
Record $15.84B quarterly revenue
Sources: Microsoft and Amazon Q1/Q3 2026 earnings disclosures; Cisco Q3 FY2026 earnings call.
Microsoft: record revenue, Xbox gutted
Microsoft’s fiscal Q3 2026 numbers, reported April 29, were about as strong as a quarter gets: $82.89 billion in revenue, up 18% year over year, with net income jumping to $31.78 billion from $25.82 billion a year earlier. The company also guided full calendar-year 2026 capex to roughly $190 billion, a 61% jump from 2025 and well past what Wall Street had modeled.
The same quarter, Microsoft cut 4,800 jobs, most of them in the Xbox gaming division, on top of 9,100 roles eliminated about a year earlier. Chief people officer Amy Coleman told staff the cuts weren’t direct AI replacements, even while acknowledging AI is reshaping how the company runs. Microsoft also rolled out its first-ever voluntary buyout program, open to senior director level and below with enough age plus tenure to qualify.
Here’s the part that undercuts the simplest version of the story: Xbox isn’t where Microsoft’s AI money is going. The division that got hit hardest wasn’t competing for capex dollars with Azure’s AI buildout in any direct sense. It just wasn’t the priority, and priority is what actually decides who keeps their job in 2026, not whether AI can technically do the work.
Amazon: 30,000 gone, $200 billion committed
Amazon’s Q1 2026 results, also reported April 29, delivered $181.5 billion in revenue and a record 13.1% operating margin, the highest in the company’s history. AWS grew 28% year over year to $37.6 billion, its fastest growth rate in 15 quarters. Amazon reiterated guidance toward roughly $200 billion in full-year 2026 capex.
Against that backdrop, Amazon cut around 30,000 corporate jobs across rounds in October 2025 and January 2026, with further cuts hitting Selling Partner Services staff and a temporary Homestead, Florida warehouse closure eliminating 600-plus jobs between July and September.
Unlike Microsoft and Cisco, Amazon’s leadership hasn’t tried to soften the connection. CEO Andy Jassy told staff in a memo, later reiterated into 2026, that generative AI and agents would reduce the company’s total corporate workforce over time as efficiency gains materialize, alongside creating new roles elsewhere. That memo dates to July 2025, a full year before this round of cuts, which makes it less a same-day justification and more a stated multi-year strategy Amazon is now executing on schedule.
Cisco: “not a savings-driven restructure”
Cisco reported record quarterly revenue of $15.84 billion on May 13, up 12% year over year, alongside AI infrastructure orders of $2.1 billion that quarter and $5.3 billion cumulative through three quarters. That pushed Cisco to raise its full-year AI order guidance from $5 billion to $9 billion, roughly 4.5 times fiscal 2025’s total.
The same week, Cisco began notifying employees that it would cut fewer than 4,000 jobs, about 5% of its global headcount, with restructuring costs running as high as $1 billion, mostly severance. CFO Mark Patterson gave analysts a line worth sitting with:
“This was really not a savings-driven restructure.”
Mark Patterson, CFO, Cisco Systems, Q3 FY2026 earnings call, via Yahoo Finance
Patterson framed the cuts as a realignment toward silicon, optics, security, and AI rather than a cost play. That’s a notably different posture from Amazon’s Jassy, and it matters: two companies profiled in the same story, cutting staff in the same season, and disagreeing with each other about whether AI is even the reason.
Is AI actually the reason, or the excuse?
Not everyone buys the AI-driven narrative, and the skepticism comes from serious places.
Layoffs are often just standard cost-cutting with an AI label attached.
Paraphrased position of Justin Wolfers, Professor of Economics and Public Policy, University of Michigan, via Benzinga/Finviz
Wolfers argues AI functions as a convenient cover story for restructuring that companies would likely have pursued regardless. JPMorgan’s 2026 economic outlook backs that skepticism with data: the bank’s own labor-market analysis found the AI capex surge hasn’t shown much measurable impact on broader labor dynamics, despite the headlines.
Wall Street’s bull case sees it differently. Wedbush’s Dan Ives, writing about Meta’s own 8,000-role cut against $135 billion in AI capex, called the layoffs financially minor next to the infrastructure commitment, damaging to morale but not decisive to the balance sheet. Evercore ISI’s Mark Mahaney goes further, noting this pattern of workforce actions followed by 12 to 18 months of margin expansion has repeated roughly every one to two years since 2022. In his read, this isn’t new behavior. It’s a recurring capital-discipline cycle that happens to be colliding with an AI narrative people want to believe.
Our read: both things are probably true at once. AI capex is real, historically large, and reshaping where investment goes inside these companies. But the specific decision to cut a specific team often has more to do with which function sits outside this year’s priority list than with any AI system actually replacing a job. Xbox wasn’t cut because a model can ship games. It was cut because it wasn’t where the $190 billion was going.
Worth flagging: Tracking firms don’t agree on the scale of 2026’s layoff wave. Layoffs.fyi-based counts put tech layoffs past 100,000 by early May and over 165,000 by July. SkillSyncer’s broader tracker counts 205,832 people affected across 322 events as of July 22, with 54% of those events explicitly citing AI or automation. The methodologies differ (tech-only versus all-industry, corporate versus contractor), so treat any single total as one tracker’s view, not a consensus figure.
What this means if you work there
If you’re an engineer or manager inside one of these companies, or one like them, the record-revenue headline is not protection. Internal budget decisions are increasingly decoupled from how well your specific team is performing. A well-run division can still get cut if it sits outside whatever core the company is funding this year, silicon, optics, security, and AI at Cisco; cloud and AI infrastructure at Microsoft and Amazon.
The more useful signal than “is the company doing well” is “where is the capex actually going.” Read the earnings call transcript, not just the headline. That’s where you find out whether your function is this year’s priority or this year’s line item.
There’s a real hedge here too. Industry-wide, roughly 275,000 AI-related roles are sitting open while laid-off tech workers largely can’t cross the skills gap to fill them, according to an Invezz analysis of labor-market data. Internal mobility toward AI or infrastructure teams is a legitimate near-term move, but it requires demonstrable fluency, not just tenure at the company.
If you’re evaluating these companies as a vendor rather than an employer, the same logic applies from the other side. Support and account-management staff, the exact functions cut at Amazon’s Selling Partner Services, may thin even as infrastructure capacity grows. That’s worth a line in any vendor risk review.
Frequently asked questions
Why are Microsoft, Amazon, and Cisco laying off workers if their revenue is at record highs?
Revenue and layoffs aren’t directly linked. Each company is redirecting tens of billions toward AI infrastructure capex, roughly $190 billion at Microsoft and $200 billion at Amazon for 2026, while separately restructuring specific divisions that sit outside their AI and cloud growth priorities.
How much is Big Tech spending on AI infrastructure in 2026?
Google, Amazon, Meta, and Microsoft combined are projected to spend roughly $725 billion on AI capital expenditure in 2026, up about 77% year over year, according to aggregated company guidance.
Is AI actually causing the 2026 tech layoffs?
It’s contested. Economists including Justin Wolfers argue AI often serves as a convenient explanation for ordinary cost-cutting. JPMorgan’s own analysis found little measurable labor-market impact from the AI capex surge, even as companies cite AI in layoff announcements.
Did Cisco lay off workers despite good earnings?
Yes. Cisco posted record Q3 FY2026 revenue of $15.84 billion, up 12% year over year, and in the same week announced plans to cut nearly 4,000 jobs as part of a restructuring its CFO described as not savings-driven.
How many tech layoffs have there been in 2026?
Estimates vary by tracker. Layoffs.fyi-based counts show over 165,000 tech layoffs by July 2026, while SkillSyncer’s broader tracker counts 205,832 people affected across 322 events as of July 22, with 54% of events citing AI as a factor.
Where this goes next
What’s changed by walking through all three companies together is this: the “AI is taking jobs” framing is too simple, and so is “it’s just normal cost-cutting.” What’s actually happening is a capital reallocation on a scale large enough that headcount decisions have become almost a separate conversation from infrastructure decisions, loosely connected at best, openly denied at Cisco, openly claimed at Amazon.
Watch three things over the next 6 to 18 months. First, whether Microsoft’s own admission that it will remain capacity-constrained through 2026 even after this spending turns into visible AI revenue, or into a monetization lag that makes the capex look premature. Second, whether more executives start talking like Jassy (AI explicitly reducing headcount) instead of like Patterson (AI reorganizing headcount). Third, whether policymakers, following California’s move this June to build a state tracking tool for AI’s workforce impact, start requiring the kind of capex-versus-headcount disclosure that would make stories like this one unnecessary.
None of these companies are lying when they post record revenue. None of them are lying when they cite AI in a restructuring memo either. They’re just optimizing for two different things at once, and reading the earnings call is currently the only way to tell which one is driving a specific decision.
Want the next capex disclosure and layoff filing broken down like this one? Subscribe to The Neural Loop at neuralwired.com/newsletter.
Your engineering team just spent $4.82 running Claude Opus 4.8 on a routine bug fix that a $0.07 model would have solved just as well. That’s not a hypothetical. It’s the real spread Artificial Analysis measured on its Coding Agent Index this year, and it’s the single most important fact in the best AI models for agentic coding tasks 2026 conversation right now. Model choice used to be about which one scored highest. In 2026, it’s about which one earns its price on the specific task in front of you.
That shift didn’t happen quietly. Six weeks ago, one of the most capable coding models on the market vanished overnight because of a U.S. export control order, then came back three weeks later. Vendors quietly stopped reporting the benchmark everyone used to trust. And developers, according to a JetBrains-backed survey, now spend more hours reviewing AI-written code than writing it themselves. This piece walks through what’s actually true, what’s marketing, and which model belongs on which job.
Why the old benchmarks stopped telling the truth
For most of 2025, SWE-bench Verified was the number everyone quoted. Scores climbed from single digits to the high 80s and low 90s in under two years, a curve that looked like genuine progress until you asked the obvious question: how do models keep getting smarter at solving GitHub issues that were published years before their training cutoff?
In February 2026, OpenAI’s own Frontier Evals team answered that question by walking away from the benchmark entirely. Their reasoning was blunt: model training had absorbed enough of the dataset that the score stopped measuring skill on unseen code and started measuring memorization. An independent audit of the top 30 leaderboard entries found that roughly 19.78% of cases labeled “solved” were passing unit tests by coincidence or by gaming the evaluation harness rather than by producing correct code.
That’s why serious 2026 comparisons have moved to two newer references: SWE-bench Pro, built on private, professional repositories that no model has seen in training, and Terminal-Bench 2.1, which scores the model and its coding harness together as they complete a real terminal-driven task from start to finish. If a vendor is still leading its marketing with a SWE-bench Verified score above 90%, read it the way you’d read a car’s mileage sticker before the EPA got involved.
The 2026 lineup, ranked
Here’s where the six models actually land once you strip out the marketing and look at SWE-bench Pro and Terminal-Bench 2.1, the two benchmarks least contaminated by memorization.
Model
Vendor
SWE-bench Pro
Terminal-Bench 2.1
Pricing (input/output per MTok)
GPT-5.6 “Sol”
OpenAI
Not separately reported
88.8% (highest recorded)
Not disclosed at review time
Claude Fable 5
Anthropic
80.3% (leader)
83.1% (Claude Code)
$10 / $50
Claude Opus 4.8
Anthropic
69.2%
78.9% (Claude Code)
$5 / $25
GPT-5.5
OpenAI
58.6%
83.4% (with Codex)
Not disclosed at review time
Gemini 3.5 Flash
Google DeepMind
55.1%
76.2%
Not disclosed at review time
Grok 4.5
xAI
Not separately reported
Not separately reported
$2 / $6
Two things jump out. First, Claude Fable 5 leads the harder, contamination-resistant benchmark by a wide margin, 11 points ahead of Anthropic’s own Opus 4.8. Second, GPT-5.6 Sol leads the benchmark that best reflects how a coding agent behaves in an actual terminal, doing real multi-step work rather than generating a single patch. Neither model is the “best” one. They’re the best at different jobs.
“But the improvement I keep coming back to is honesty.”
Rahul Patil, CTO, Anthropic, on Claude Opus 4.8’s jump on SWE-bench Pro, via EdTech Innovation Hub
Patil described the target workload for Opus 4.8 as the kind of job that “used to take a quarter and a working group,” meaning codebase-scale migrations and bug fixes spread across hundreds of files. That framing matters. It’s a tacit admission that raw benchmark points matter less than whether the model can survive a genuinely large, messy, real-world job without losing the thread.
Where the open-weight tier fits in
Not every team needs frontier pricing. GLM-5.2 from Z.ai, released under an MIT license, scores 62.1 on SWE-bench Pro and 81.0 on Terminal-Bench 2.1 at $1.40/$4.40 per million tokens, the strongest published open-weight coding numbers available right now. Kimi K2.6 Code lands an 80.2% SWE-bench Verified score at open-model pricing, close to Opus-class accuracy for a fraction of the bill. Neither will win a head-to-head against Fable 5 on the hardest tasks. Both will handle the routine 80% of your ticket queue for pennies.
What each task actually costs you
Here’s the number that should reshape how your team budgets for AI coding tools: on Artificial Analysis’s Coding Agent Index, an identical fixed task costs about $0.07 to run through Cursor’s Composer 2.5 and $4.82 to run through GPT-5.5, a roughly 60x spread for only three to four points of quality difference on the index.
The mechanism most teams miss
Coding agents burn most of their budget on reading, not writing. PointFive’s July 2026 index found that a single realistic task, reading a handful of files, reasoning about them, and revising a diff, can pull in up to 200,000 input tokens against a diff of roughly 30,000 tokens written back. That’s why input pricing, and whether a model caches repeated reads at a discount (often around 10% of the standard rate), swings your real bill far more than the headline output price per token.
Zoom out and the frontier-to-workhorse spread gets even wider. Claude Fable 5 charges $10/$50 per million tokens. DeepSeek V4 Flash charges roughly $0.14. That’s close to a 180x difference in raw token pricing between the most expensive and cheapest options a team might reasonably put in production this year.
None of this means cheap wins by default. GLM-4.6 costs about $0.059 per task but is statistically tied on accuracy with pricier open options, which means the math sometimes favors a marginally more expensive model like DeepSeek or Qwen instead. The lesson isn’t “buy the cheapest model.” It’s “stop assuming the most expensive model is the safest default,” and start routing tasks by difficulty: cheap model first, escalate to a frontier model only when the cheap one fails.
The Fable 5 warning every team should have caught
Claude Fable 5 launched June 9, 2026, and immediately topped the SWE-bench Pro leaderboard. Three days later, on June 12, 2026, Anthropic suspended it worldwide to comply with a U.S. Department of Commerce export control order. Access came back on July 1, 2026, after the controls were lifted, and Anthropic confirmed the restoration directly.
Three weeks of downtime for a model teams were actively shipping in production. If your pipeline depended entirely on Fable 5 during that window, you didn’t have a benchmark problem. You had a supply chain problem, and most engineering leaders still aren’t tracking it as one.
There’s a second wrinkle independent evaluators caught after Fable 5 came back online: Artificial Analysis and Vals AI both measured Fable 5 refusing roughly 8 to 9% of test prompts, quietly falling back to Opus 4.8 for those cases. That means the headline SWE-bench Pro score doesn’t fully describe what a production deployment experiences. A meaningful slice of real traffic never actually touches the model you thought you were paying for.
Best practice going forward: never build a single-model dependency into a critical pipeline. Keep at least one fallback model configured, and treat a vendor’s top model the way you’d treat a single-region cloud deployment. It works great, right up until it doesn’t.
The bottleneck nobody’s marketing deck mentions
Every vendor above is racing to add benchmark points. Almost none of them are talking about the actual reason enterprise AI coding adoption stalls, and that’s reliability, not raw capability.
“It unpacks different factors that I see tangled together in almost every eval I’ve ever seen.”
Bryan Silverthorn, Director of AGI Autonomy, Amazon, at VB Transform 2026
Silverthorn, who joined Amazon through its Adept AI acquisition, argues that “reliability” isn’t one thing. He breaks it into four separate dimensions, borrowing a framework from Princeton research: consistency, robustness, predictability, and safety. He described a customer whose agent performed a serial number extraction task flawlessly for two months, then quietly started misreading numbers with no warning and no obvious trigger. No benchmark on this list would have caught that failure mode before it hit production.
Paul Gauthier, creator of the open source pair-programming tool Aider, has built a reputation on the opposite end of the spectrum: refusing to rank his own tool against agents that won’t publish their evaluation methodology. If a vendor won’t show its work, that’s a signal worth weighing as heavily as the score itself.
There’s a human cost showing up in the data too. A developer survey compiling adoption research found that engineers using AI coding tools now spend 11.4 hours a week reviewing AI-generated code, against 9.8 hours writing new code themselves, a reversal from the pattern two years ago. The “10x productivity” pitch quietly assumes review time is free. It isn’t.
How to actually choose, task by task
Stop asking which model is smartest. Ask which model fits the task type sitting in your queue right now.
CLI-heavy DevOps and multi-step terminal work: GPT-5.6 Sol currently leads Terminal-Bench 2.1 at 88.8%, the strongest publicly reported score for real terminal-agent workflows.
The hardest multi-file repository repairs: Claude Fable 5 leads SWE-bench Pro, provided you’ve built in a fallback for its 8 to 9% refusal rate and you’re comfortable with the export control volatility above.
Large-scale migrations and refactors across hundreds of files: Claude Opus 4.8, purpose-built by Anthropic for exactly this workload, with its Dynamic Workflows feature fanning work out to parallel subagents.
Tool-orchestration-heavy agent work: Gemini 3.5 Flash leads MCP Atlas at 83.6% even though it trails on raw SWE-bench numbers, making it a genuine specialist pick for agentic tool-calling.
The routine 80% of your ticket queue: An open-weight model like GLM-5.2 or Kimi K2.6, or a workhorse like Cursor’s Composer 2.5, saves 10 to 60x on cost for a 3 to 4 point accuracy trade-off most teams won’t even notice.
Our read: the real 2026 skill isn’t picking a single model and standardizing on it. It’s building a routing layer that sends each task to the cheapest model likely to solve it, and escalates only on failure. Teams still budgeting per seat instead of per completed task are leaving real money on the table, and the PointFive and Artificial Analysis data above shows exactly how much.
Frequently asked questions
What is the best AI model for coding in 2026?
There’s no single winner. Claude Fable 5 leads the hardest contamination-resistant benchmark, SWE-bench Pro. GPT-5.6 Sol leads real terminal-agent work, scoring 88.8% on Terminal-Bench 2.1. The right choice depends on task type and budget, and open-weight models like GLM-5.2 close most of the gap at a fraction of the cost.
How much does an AI coding agent cost per task?
Cost per completed coding task ranges from roughly $0.07 to $4.82 depending on the model, according to Artificial Analysis and PointFive benchmark data. Workhorse models like Cursor’s Composer 2.5 cost around $0.07 per task, while frontier models like GPT-5.5 or Claude Opus can run $4 or more for only a few extra benchmark points.
Why did OpenAI stop reporting SWE-bench Verified scores?
OpenAI’s Frontier Evals team announced in February 2026 that it would stop reporting SWE-bench Verified results because training data contamination had inflated scores past the point where they reflected real coding ability on unseen code. SWE-bench Pro, built on private repositories, is now the more trusted reference.
Is Claude Fable 5 still available?
Yes. Claude Fable 5 launched June 9, 2026, was suspended worldwide on June 12, 2026 under a U.S. Department of Commerce export control order, and access was restored on July 1, 2026 after the controls were lifted. Teams building on it should keep a fallback model plan in place given that volatility.
