⚡Anthropic grants outside evaluator METR employee-level access to its AI models for independent safety evaluation.

What it means

Anthropic is giving six researchers from METR, an independent nonprofit that evaluates frontier AI models, employee-level access to its internal operations—including company hardware, office space, and model access—so they can conduct independent safety assessments and publish findings without prior approval. This is a notable step toward external oversight in AI safety, aiming to increase transparency and accountability beyond typical third-party audits.

Accuracy & sourcesSourced

The claim is well-supported by multiple credible news outlets (TechCrunch, Reuters, CNBC) reporting on official Anthropic statements, plus a direct blog post from Anthropic. The primary sources are Anthropic's own update and METR's role as the independent evaluator, so the core fact is solid.

How new