Technology and Scienceupdated 1h ago2 min read
Anthropic Grants METR Employee-Level Access to Assess AI Risks, Will Publish Results
This article is an automated aggregation of coverage from news outlets. Our editorials offer in-depth analysis curated by the NewsCord team.
Six METR researchers granted employee-level access to Anthropic's internal systems for independent evaluation. They will receive equipment, office space, and access to internal tools comparable to risk-assessment team.
10 outlets told this the same way.
METR gets employee-like access
METR receives employee-like access to a team from METR, a nonprofit that evaluates frontier AI models, and METR will publish the results of its risk assessments. METR employees receive company hardware, office space, and access to internal tools that are roughly on par with what its own risk assessment teams have access to. Anthropic signed the White House Accord on Super Intelligence, in which companies pledged to partner with independent external auditors or evaluators.
Anthropic is working with SecureBio and Accenture as evaluators and is in early discussions with several other organizations, including Hugging Face. Anthropic wants to close the gap between what evaluators understand about a frontier AI company’s systems and what the company itself understands.
Publication without veto, limited redactions
METR will publish its analysis of its risk levels, incidents, or practices, and Anthropic cannot veto publication but can require redactions in narrow, predefined categories. METR is free to characterize its redactions publicly and say whether it believes the redactions were justified, as long as it does so without revealing the redacted information. Anthropic’s engagement with METR is limited to a small team operating under enforceable agreements, and other METR employees are siloed from that team’s work.
Anthropic is developing three tiers of access for evaluators, with the first tier allowing narrowly scoped risk investigations and publication with Anthropic’s approval. The second tier allows evaluators to publish without Anthropic’s approval while Anthropic can still require removal of specific categories of information.
Why the access model matters
Anthropic compared its approach to bank supervision, saying regulators assign resident examiners to work inside the largest banks when a bank’s risk-taking could threaten the public. Anthropic cannot publicly share all the information relevant to its safety work because some of it could be useful to malicious actors or is commercially sensitive, like compute or training information. TechCrunch said METR is the only evaluator so far at the top of a three-tier system Anthropic is building, with SecureBio in the bottom tier running narrow reviews that require Anthropic’s approval to publish.
Anthropic wants evaluators to have expertise in assessing the capabilities of frontier models, independence from financial ties, enough access to meaningfully evaluate risks, and freedom to publish observations and assessments. Anthropic is moving quickly with multiple evaluators and learning actively about what kinds of partnership work it can support.