EU cyber agency ENISA to test Anthropic's Mythos AI scanner

The European Union Agency for Cybersecurity, ENISA, is set to gain access to Anthropic's advanced AI vulnerability scanner, Claude Mythos, according to sources familiar with the matter. The arrangement would make the EU's top cyber authority the first entity outside the United States and United Kingdom to test the powerful model through Anthropic's Project Glasswing initiative.

Project Glasswing is a controlled-access program launched in April, designed to allow a select group of organizations to use Mythos to identify security flaws in their systems before a wider release. The initial roster of roughly 50 partners includes major technology firms such as Amazon, Apple, Microsoft, Google, Nvidia, and cybersecurity companies like CrowdStrike and Palo Alto Networks, as well as financial institutions.

The decision follows discussions between Anthropic and officials from the European Commission, the EU's executive branch. Commission officials had traveled to San Francisco to negotiate access after the model was previewed. A spokesperson for ENISA confirmed that an offer had been made, but stated that conditions were still being agreed upon. A European Commission spokesperson noted...