Anthropic gives outside AI watchdog “employee-like” access
The outside watchdog can publish its findings without Anthropic’s approval.
TL;DR
- Anthropic is giving METR, a nonprofit, extensive access to its systems and employees to evaluate AI model risks.
- This move is intended to increase safeguards and public accountability for powerful AI technologies.
- The arrangement allows METR to publish its findings without Anthropic's veto, though certain data redactions are permitted.
- Anthropic sees this as a model for other frontier AI companies, drawing comparisons to bank supervision.
- The company is also working with Accenture and SecureBio on safety testing, and is under pressure to reassure investors about safety ahead of a potential IPO.