Anthropic invites EU cybersecurity agency to access Mythos AI hacking model

Anthropic has invited the European Union to access Mythos, its powerful AI-powered hacking tool, by sending an invitation to the bloc’s cybersecurity agency.

A European Commission official said the AI company issued the formal invitation following a meeting with the Commission in San Francisco last Thursday, adding that the EU must now establish a mechanism that would allow access to the model under appropriate security safeguards.

Bloomberg reported on Monday that ENISA, the EU’s cybersecurity agency based in Athens, would be granted access to Mythos.

European Commission spokesperson Thomas Regnier said the Commission had held “several productive meetings with Anthropic” and “welcomes the latest developments regarding potential future access.”

Anthropic unveiled Mythos in early April and warned that the model outperformed most humans in identifying and exploiting cybersecurity vulnerabilities.

The disclosure raised concerns that the model could be used to carry out large-scale attacks against critical and sensitive systems if it fell into the hands of cyber adversaries.

European officials were unable to access the cutting-edge cybersecurity AI technology for weeks, prompting urgent calls from European lawmakers and government officials to secure access.

Cybersecurity officials also urged Europe to develop its own version of the technology.

“This latest development is extremely important in helping us gain a clear understanding of the potential risks. We should not forget that Mythos is not an isolated case and that a new wave of powerful models is entering the market,” Regnier said.

An ENISA official said the agency does not currently have active access to the model but is working to make it operational.

The Commission is developing a formal action plan to respond to powerful AI hacking tools.

According to an industry official, the Commission has indicated that it wants to publish the plan before the summer break.