Anthropic grants EU cybersecurity agency access to Mythos AI model

The European Union Agency for Cybersecurity, known as ENISA, has become the first EU body to gain access to Anthropic’s Mythos AI model. Testing began on September 10, 2026, following months of negotiations that were complicated by US export controls on advanced AI technologies. The access agreement positions Europe’s primary cybersecurity authority to evaluate what Anthropic describes as a defensively oriented AI system, one that has already demonstrated some eyebrow-raising capabilities during earlier testing phases with other partners. What Mythos actually does Mythos is primarily designed for defensive cybersecurity applications. During preliminary testing, the model autonomously identified over 10,000 zero-day vulnerabilities. Zero-days are software flaws that vendors don’t yet know about, the kind of weaknesses that nation-state hackers and criminal groups pay millions to exploit. The model also reportedly achieves success rates exceeding 83% in generating effective exploits on its first attempt. That’s a significant leap over previous models like Opus 4.6. The ENISA arrangement grew out of Project Glasswing, Anthropic’s structured rollout program for Mythos access. That proj...