Anthropic empowers external evaluators with enhanced access

Anthropic has granted external evaluators, including the nonprofit METR, 'employee-like' access to its systems to assess the risks associated with its AI models. This move is part of the company's effort to enhance safeguards against potential catastrophic risks from its technology. METR, which has previously collaborated with major AI labs such as OpenAI, Google, and Meta, now has a team embedded within Anthropic's office, providing them with access to hardware and internal systems. This increased involvement of third-party evaluators is one of the key steps proposed by AI labs to address growing concerns about the potential dangers of more advanced AI systems to the economy, cybersecurity, and global security. The decision reflects a broader trend among leading AI organizations to prioritize transparency and external oversight. By allowing external entities to scrutinize their operations, companies like Anthropic aim to build trust and mitigate risks associated with the rapid development of AI technologies.