Anthropic gives METR researchers employee‑level access
anthropic
| Source: TechCrunch | Original article
Anthropic gave METR researchers employee-level access, letting them independently assess the company's risks, incidents and practices without Anthropic's right to block the review.
Anthropic has granted the Model Evaluation and Transparency Research (METR) group employee‑level access to its internal systems, allowing the independent team to publish its own assessments of the company’s risks, incidents and operational practices. Crucially, Anthropic has waived its usual right to block or edit those assessments, opening a rare window into the day‑to‑day workings of a frontier AI developer.
The move arrives as regulators and investors intensify scrutiny of large‑scale AI firms. Earlier this month the FTC announced a probe into Anthropic’s practices, and investors have already signalled a downward shift in the company’s valuation expectations. By inviting METR to examine its code, data pipelines and incident logs, Anthropic is attempting to demonstrate transparency and pre‑empt further regulatory pressure. For the broader AI ecosystem, the arrangement could set a new benchmark for third‑party oversight, offering a model for how companies might voluntarily expose internal safeguards without compromising proprietary technology.
What to watch next: METR’s first report, slated for release later this year, will reveal how Anthropic’s risk‑management processes stack up against industry standards. Stakeholders will be looking for concrete findings on model safety, data handling and incident response, as well as any recommendations for improvement. Regulators may cite the findings in ongoing investigations, while investors could factor the transparency initiative into future financing rounds or IPO plans. The industry will also be watching whether other AI labs adopt similar open‑access arrangements, potentially reshaping norms around accountability in the fast‑moving frontier AI sector.
Sources
Back to AIPULSEN