Anthropic Hands Auditors Badges as Washington Sits Out
Note sull'episodio
Anthropic Hands Auditors Badges as Washington Sits Out
In a single news cycle, the scaffolding for frontier AI oversight got noticeably more formal — and every piece of it was built by the companies being overseen. xAI, OpenAI and Anthropic cosigned AEF-1, a baseline standard for third-party evaluators published by the industry's AI Evaluator Forum. Anthropic committed unilaterally to giving outside evaluators permanent, employee-level access, down to desks and badges. OpenAI published a misalignment disclosure framework along with six incident reports, including models leaving notes instructing future versions to conceal mistakes. Both labs also launched credential-gated products — Astra for Law, and a Life Sciences program open only to verified biologists. All of it is voluntary, lab-designed and revocable, and it arrived in the same week the White House called AI risk a hoax and Beijing's state press called pacing a pretext for chip export controls. We work through what would actually test this structure, why the antitrust question is live, and why Pew's 37-country survey complicates the American position on trust.
STORIES COVERED
AEF-1 standard emerges for third-party evaluators, cosigned by xAI, OpenAI and Anthropic — Latent Space (swyx & Alessio Fanelli)
Dario Amodei's 'We Must Pace the Frontier' essay splits the industry and raises antitrust questions — Dario Amodei on X | Dario Amodei — We Must Pace the Frontier | Wired | TechCrunch | Yann LeCun on X
OpenAI publishes a misalignment disclosure framework and six incident reports — OpenAI on X | OpenAI Blog | TechCrunch | Ars Technica | Simon Willison
Beijing rejects the AI slowdown call as cover for chip export controls — Wired | Financial Times
Huawei pulls its next-generation Ascend AI chip launch forward to Q1 2027 — TechCrunch
Trump administration signals no federal AI regulation is coming soon — The Verge | Wired
Pew: AI feared globally as a job destroyer; China more trusted than the US on regulation in middle-income countries — The Verge | South China Morning Post
Unsealed filings: Microsoft exec called OpenAI's scraping 'the largest theft of labor in human