Anthropic gives EU agency access to its cybersecurity model Mythos

AI company Anthropic is planning to give the European Union Agency for Cybersecurity (ENISA) access to its cybersecurity model Mythos, so the agency can test it before a wider release. The decision was shared with the European Commission last weekend.

Since Mythos launched in late April, its powerful ability to spot cybersecurity weaknesses has drawn attention. EU officials have been pushing for access, and last week they traveled to San Francisco to meet with Anthropic's executives.

Anthropic says Mythos could pose serious risks, so it's taking a gradual approach, letting governments and companies test system security first. EU spokesperson Thomas Regnier said more powerful models are coming, and the EU is discussing challenges with partners like the US.