Anthropic puts three evaluator access tiers in writing, with METR at the top
Anthropic's new framework gives six embedded METR researchers employee-like access and a right to publish without approval. Tech Times said it followed a Reuters report that Anthropic sought strict editorial authority over evaluator reports as late as Oct. 5.
Reality
- Evidence40
- Adoption25
- Hype gap+15
- Incentives55
- Confidence45
buildOne report1 publisher Anthropic puts a six-person METR team inside its offices and internal tools
Anthropic will give a six-person METR team employee-like access to its hardware, offices and internal tools. Employees can share anything with it except customer data, privileged legal material and model-architecture details.
Reality
- Evidence35
- Adoption
- Not scored
- Hype gap+10
- Incentives60
- Confidence50
Workday's Rob Enslin said the C1.ai acquisition makes Workday the identity governance layer for people, money and agents. The customers on stage, BMO and Intermountain Health, framed their own agent controls as scope, spending and sign-off.
Reality
- Evidence45
- Adoption40
- Hype gap+25
- Incentives70
- Confidence60
Researchers say AI ethics critics treat several different goals as one
Researchers at Bayreuth, Saarland and CISPA sort AI ethics into three objectives and four domains in a paper in AI and Ethics. They argue that criticism of one instrument, such as company guidelines, should not automatically call the whole field into question.
Reality
- Evidence30
- Adoption
- Not scored
- Hype gap−5
- Incentives40
- Confidence55
Fewer than half of 1,100 leaders have the integrated risk model two-thirds call a priority
In Grant Thornton's survey of 1,100 business leaders, 67% rank enterprise risk management a top priority and 48% have implemented an integrated model. Separately, 61% say insufficient integration across risk functions leaves an incomplete view of organizational risk.
Reality
- Evidence40
- Adoption48
- Hype gap+15
- Incentives70
- Confidence50
IFS survey finds just 4% of organisations mandate human approval of AI-supported decisions
IFS's survey of more than 2,200 senior decision makers in 14 countries found only 4% of organisations require human approval of AI decisions. IFS published it beside its own AI platform. The gap it describes is also its sales case.
Reality
- Evidence35
- Adoption30
- Hype gap+40
- Incentives85
- Confidence55
67% of AI-using small and mid-sized firms run tools they built, Pax8 survey finds
Pax8's survey of 1,150 business leaders finds 67% of AI-using small and mid-sized firms have built an AI tool now in regular use. In every market most builders lack a formal, documented review, so governance, security and cost questions fall to the IT providers who serve them.
Reality
- Evidence45
- Adoption55
- Hype gap+15
- Incentives70
- Confidence50
Gartner forecasts over 40% of agentic AI projects canceled even as a third of software ships with agents
Gartner expects over 40% of agentic AI projects to be canceled by the end of 2027 and a third of enterprise software to ship with agentic AI by 2028. Quartz's summary says both can hold at once. We think that argues for releasing agent budgets in tranches.
Reality
- Evidence35
- Adoption
- Not scored
- Hype gap
- Not scored
- Incentives
- Not scored
- Confidence40
buildConfirmed5 publishers Anthropic says normal safeguards were off and the network was live during third-party tests, and that its own pre-release auditing gave no warning. METR now gets at least eight weeks of broad access.
Perspective Coverage
5 publishers
- Builder
- Builder 45%
- Operator
- Operator 36%
- Investor
- Investor 19%
Reality
- Evidence66
- Adoption
- Not scored
- Hype gap+10
- Incentives58
- Confidence64
President Trump named intelligence director Jay Clayton to lead a White House AI task force with 120 days to set a federal plan for AI risks. With the spy chief in charge, AI companies are likelier to face federal demands on incident reporting than rules on how models are built.
Perspective Coverage
39 publishers
- Builder
- Builder 28%
- Operator
- Operator 41%
- Investor
- Investor 31%
Reality
- Evidence72
- Adoption
- Not scored
- Hype gap+30
- Incentives62
- Confidence66
buildOne report1 publisher TypeSafe's Jev, Cloudflare's Clef and AWS's Strands Decider 2B bring small models to agent decisions
Cloudflare and AWS have followed TypeSafe's Jev with Clef and Strands Decider 2B, small models built to pick an agent's next tool or step. The pitch is a smaller token bill, though the practitioners dev.to cites say calibration and the cost of a wrong pick decide whether any saving survives.
Reality
- Evidence30
- Adoption
- Not scored
- Hype gap+25
- Incentives
- Not scored
- Confidence35
ISACA opens beta applications for an AI governance certification due in early 2027
ISACA is taking beta applications for an AI governance certification it plans to launch in early 2027. The gap it is selling into comes from ISACA's own polling, where only 32% of digital trust professionals say their employers handle AI risk adequately.
Reality
- Evidence40
- Adoption
- Not scored
- Hype gap+20
- Incentives80
- Confidence45
PCI SSC asks for human approval before AI agents act on cleartext card data
PCI SSC's new guidance on AI systems recommends explicit human approval for any agent action involving cleartext cardholder data. Its scoping rules also count an agent holding tokens plus a detokenization tool as an agent that can read card data.
Reality
- Evidence55
- Adoption
- Not scored
- Hype gap+5
- Incentives
- Not scored
- Confidence55
Companies planning autonomous IT operations still want a person to approve what AI does
Ninety percent of companies surveyed want autonomous IT operations within two years, yet 77% hesitate to let AI decide without human sign-off. The same respondents' readiness and visibility figures show most lack the groundwork even for agents that act under supervision.
Reality
- Evidence35
- Adoption20
- Hype gap+25
- Incentives
- Not scored
- Confidence35
China has shipped about half the 25 million tons of soybeans it promised, and the suspension of its rare earth export controls runs out on Nov. 10. Bank of America's base case is a one-year extension.
Perspective Coverage
10 publishers
- Builder
- Builder 21%
- Operator
- Operator 32%
- Investor
- Investor 47%
Reality
- Evidence70
- Adoption45
- Hype gap+10
- Incentives65
- Confidence68
Washington and Beijing pushed their tariff truce two months past November 10 and left $30 billion a side of tariff relief at the recommendation stage. Tariff risk still has an expiry date, now around January 10, even with a new Board of Trade and formal AI talks in place.
Perspective Coverage
10 publishers
- Builder
- Builder 25%
- Operator
- Operator 38%
- Investor
- Investor 37%
Reality
- Evidence76
- Adoption30
- Hype gap+12
- Incentives58
- Confidence70
Evan Hubinger posted the number from inside the vendor on September 8, and attached to it was a written admission that Anthropic has no fix for superintelligence alignment.
Perspective Coverage
3 publishers
- Builder
- Builder 23%
- Operator
- Operator 42%
- Investor
- Investor 35%
Reality
- Evidence62
- Adoption
- Not scored
- Hype gap+20
- Incentives45
- Confidence65
Both capitals say slowing AI development is unnecessary, so the concrete item going into Wednesday's talks is a bilateral notification channel for AI incidents, with the next treasury session two months out.
Perspective Coverage
3 publishers
- Builder
- Builder 20%
- Operator
- Operator 48%
- Investor
- Investor 32%
Reality
- Evidence55
- Adoption
- Not scored
- Hype gap+20
- Incentives60
- Confidence50
Meta's Oversight Board tells AI labs their new safety committees need the power to cause friction
Meta's Oversight Board issued six recommendations for the safety committees six AI companies, including Google and OpenAI, pledged last week. The accord behind the pledge would fill those committees with company directors, so the independence the board calls essential is in doubt from the start.
Reality
- Evidence55
- Adoption
- Not scored
- Hype gap+5
- Incentives50
- Confidence60
For buyers of American AI, Trump's new czar matters more than the CEOs' self-policing pact
Donald Trump signed a 'morally binding' AI pact with six CEOs, then made intelligence chief Jay Clayton his AI czar. For firms building on American models, the appointment carries more consequence than a pledge they have no way to enforce.
Reality
- Evidence38
- Adoption
- Not scored
- Hype gap+30
- Incentives70
- Confidence35
Earlier coverage
- Appetite for agentic IT outruns the trust to let it make decisions
Leadership · October 6, 2026 · One report1 publisher
- Bank AI job postings climb 49% to 139,819 as agent orchestration grows fastest
Invest · October 2, 2026 · Confirmed2 publishers
- Frontier lab scientists ask governments to put auditors inside their own companies
Leadership · September 28, 2026 · Confirmed3 publishers
- OpenAI took about two months to learn its agents had breached an Australian Medicare-data service
Invest · September 27, 2026 · Confirmed18 publishers
- US and China agree to set up a channel for AI-related incidents
Security · September 26, 2026 · One report1 publisher
- OpenAI stops a "significant number" of Astra training runs until cyber gates are met
Product · August 18, 2026 · Confirmed7 publishers
- Gates now wants firms taxed on their AI use. Nobody has defined the base.
Invest · August 26, 2026 · Confirmed4 publishers
- Coder registers your self-hosted workspaces as a compute pool in Cursor's scheduler
Build · September 2, 2026 · Confirmed2 publishers
- Anthropic widened its dual-use biology block on lost certainty about the old threshold
Build · September 10, 2026 · Confirmed5 publishers
- Microsoft puts AI code of conduct out for public comment, with rules not guiding models until 2027
Invest · September 14, 2026 · Confirmed8 publishers
- Microsoft's AI chief attacks the training doctrine of a lab Microsoft both funds and pays
Invest · September 17, 2026 · Confirmed3 publishers
- OpenAI wrote research pace out of its new math advisory group's remit
Build · September 22, 2026 · Confirmed2 publishers
- JetBrains Air Governance extends cost and visibility tracking across AI providers
Build · September 22, 2026 · Confirmed2 publishers
- OpenAI and Anthropic asked 15 governments for AI reporting rules the White House opposes
Leadership · September 23, 2026 · Confirmed4 publishers