Anthropic has granted the European Union Agency for Cybersecurity (ENISA) access to its high-performance AI model, Mythos 5. European Commission spokesperson Thomas Regnier confirmed that the agency is currently conducting tests on the model following months of negotiations.
The Mythos model, introduced by Anthropic in April, is designed to identify cybersecurity vulnerabilities. Through the Project Glasswing initiative, access is restricted to vetted organizations to help patch security flaws before they can be exploited by malicious actors.
Negotiations between the EU and Anthropic began in late May but faced delays due to disagreements over access terms and scope. The process was further complicated by White House restrictions on foreign access to Mythos and the Fable model, though these rules were later relaxed.
Despite the agreement, the EU did not receive full access to the latest iteration of the technology. According to a European Commission spokesperson, ENISA is currently unable to access the updated Mythos 5.1 version of the model.
The UK AI Safety Institute was among the first non-US organizations to pressure-test the original Mythos model. However, sources indicate that the institute has also been denied access to the newest version of the AI system.
