Anthropic has granted the European Union Agency for Cybersecurity (ENISA) access to its advanced AI model, Mythos. The agreement follows a negotiation process that lasted over three months after the company first announced its intention to provide such access to regional authorities.
The model was introduced by Anthropic in April as part of Project Glasswing, an initiative focused on identifying and patching cybersecurity vulnerabilities before they can be exploited. The company restricts access to this technology to verified organizations to prevent misuse.
The collaboration faced significant delays due to discussions regarding the scope of access and a White House directive prohibiting foreign entities from accessing Mythos and the Fable model. While negotiations were successful, ENISA was denied access to the latest version, Mythos 5.1.
The UK AI Safety Institute was among the first non-US organizations to gain access to the original Mythos model. However, like their European counterparts, British researchers have not been granted access to the most recent 5.1 iteration of the software.
