“`html
- The UK government’s AI Security Institute has published an initial evaluation of Anthropic‘s Mythos Preview model, adding independent public verification to earlier reports suggesting its striking capabilities in cybersecurity tasks.
- This evaluation shows that Mythos isn’t significantly different from other recent frontier models in tests of individual cybersecurity-related tasks. However, it could set itself apart through its ability to effectively chain these tasks into the multi-step series necessary for fully infiltrating some systems.
- The AI Security Institute has been conducting Capture the Flag challenges since early 2023 and found that Mythos Preview can now complete over 85 percent of the group’s “Apprentice” level CTF tasks, indicating significant progress in its cybersecurity capabilities compared to previous models like GPT-3.5 Turbo.
“`
Originally published at arstechnica.com. Curated by AI Maestro.
Stay ahead of AI. Get the most important stories delivered to your inbox — no spam, no noise.