Story perspectives
ENISA first EU agency using Anthropic’s Mythos, finds 23k flaws.
6/2/2026
1 of 3
ENISA Leverages AI
- ENISA becomes the first EU agency to access Anthropic’s Claude Mythos AI model through Project Glasswing.
- The model autonomously identifies vulnerabilities and generates exploits, flagging 23,019 flaws including about 6,202 high- or critical-severity ones.
- In its first month, Mythos found over 10,000 flaws and achieved a 72.4 percent exploit success rate.
- Anthropic supplied $100 million in credits, $4 million in security donations, and named Amazon, Apple, Microsoft, Nvidia, JPMorgan Chase and Google as partners.
1 / 3
