Anthropic gives EU agency access to security AI Mythos

Anthropic has committed to granting EU cybersecurity agency ENISA access to its new language model Mythos. Handelsblatt reports that the AI tool is already being used by numerous US firms and is exceptionally skilled at identifying vulnerabilities in systems.

Strategic step for European security

Providing the model to the European authority is a strategic move: ENISA will be able to test and evaluate the capabilities and potential risks of the technology. Bloomberg reports that Anthropic describes the model as extraordinarily adept at detecting network vulnerabilities – but simultaneously warns that it could pose a significant cybersecurity threat to critical systems.

The collaboration with the EU agency underscores the growing importance of AI security in Europe. While Anthropic grants authorities access to its models, the company is simultaneously preparing for its IPO: Tagesschau reports that Anthropic – known for its AI application Claude – has already filed documents with the US securities regulator. This is the first formal step on the path to going public.

Willingness to cooperate with European regulators

The measure fits into the broader context of European AI regulation. The EU is working to monitor powerful AI systems and minimize their risks. By granting ENISA access to Mythos, Anthropic signals its willingness to cooperate with European regulatory authorities – an important signal at a time when controlling AI security risks is increasingly coming into focus.