Anthropic has granted external evaluators from non-profit METR "employee-like access" to assess the risks of its AI models, following through on a commitment to increase safeguards amid concerns about catastrophic risks. METR, which has worked with OpenAI, Google, and Meta, now has a team embedded in Anthropic’s office with access to hardware and internal systems. This move responds to growing concerns about the dangers of advanced AI systems to the economy, cybersecurity, and humanity.
Anthropic CEO Dario Amodei previously achieved rare alignment with rivals Sam Altman and Elon Musk on embedding external auditors in leading AI labs. The initiative aims to provide public accountability without exposing sensitive information that could benefit malicious actors or competitors. Anthropic argues the model resembles bank supervision, where resident examiners work inside major banks to monitor risk-taking.
Evaluators will publish their analyses, though Anthropic can require redactions for customer data, attorney-client privilege, or model architecture details. The company is not paying METR, which refuses donations from AI companies or staff and is funded by individuals, philanthropies, and government bodies like the UK’s AI Security Institute. METR previously investigated the Hugging Face hack involving rogue OpenAI models.
Anthropic has onboarded six METR employees, providing them with company hardware, office space, and access to internal tools comparable to its own risk teams. The firm is also under pressure to reassure investors ahead of an expected IPO this year at a over $2tn valuation, having emphasized safety practices in recent investor presentations.
Source: Financial Times Companies · Summarized by HeadlinesBriefing