Anthropic grants external watchdog employee-like access to its models
Anthropic announced that the nonprofit evaluator METR now has employee‑like access to its hardware and internal systems to assess model risks. METR, which has previously worked with OpenAI, Google and Meta, has a team embedded in Anthropic’s office to conduct the audits.
Key points
- Anthropic gave METR employee‑like access to its hardware and internal systems for risk assessment.
- METR, a nonprofit that has worked with OpenAI, Google and Meta, is now embedded in Anthropic’s office.
- Critics say the arrangement could create conflicts of interest and urge more government regulation.
The move follows a proposal by Anthropic CEO Dario Amodei that was supported by OpenAI’s Sam Altman and SpaceX founder Elon Musk, aiming to increase safeguards as AI models become more powerful. Anthropic said the arrangement mirrors bank supervision, where regulators place resident examiners inside major banks. Critics have argued the setup could create conflicts of interest and called for stronger government regulation. Anthropic countered that full public disclosure might aid malicious actors or competitors. OpenAI has yet to lay out plans to bring in evaluators.
The story so far
5 episodes →- Anthropic grants external watchdog employee-like access to its modelsthis story
Anthropic gives outside AI watchdog “employee-like” access
Ars Technica AI · 18 October 2026
Loading the full article…
This text was published by Ars Technica AI and written by Cristina Criddle, Financial Times. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Policy & Regulation
All →- Opinion: Canadian healthcare needs AI to keep records with patients · 1 src
- EU investigates OpenAI and Anthropic over AI agent breaches of Hugging Face and three organizations · 6 src
- OpenAI's GPT-5.6 Sol used unauthorized Sepolia ETH to deploy smart contract · 13 src
- Anthropic warns its models could self-replicate on IoT devices within months · 2 src
- Apple, Amazon and Meta could face privacy challenges from always‑listening AI wearables · 1 src
Comments
via GitHub Discussions