Safety · 2026-10-19

Anthropic Gives External Reviewers Access Comparable to Its Own Employees

Anthropic, the company that builds the Claude AI model, has given the independent organization METR access to its systems at the level of its employees so it can test them for safety.

✨ What happened

Anthropic, which develops the Claude AI model, has given METR access to its systems that is almost the same as that of its own employees. Reuters reports this, and the news is presented as part of the company's broader work on artificial intelligence safety.

The key point is that the review is carried out by an external party, not only by the developer company itself. This allows AI to be examined from an independent perspective.

📌 Who is METR

METR is an independent organization that researches and evaluates AI models. Its specialists look for whether these systems can do something dangerous or unpredictable, for example, help with harmful actions or get out of control.

For an ordinary person, this is like independent technical expertise. It is needed so that AI safety does not have to be taken on the developer's word alone.

🔹 Why it matters

As AI models become more powerful, their testing becomes more important before they are used on a massive scale. External access makes it possible to find problems that an internal team might have missed.

This news shows that the industry is gradually moving toward the practice of independent safety oversight. More detailed information, such as specific models or test results, is not included in this brief report, so it is worth following further publications.

Sources: 🔗 Google News