The UK Government’s Mythos AI Tests Help Separate Cybersecurity Threat from Hype
- The UK government’s AI Security Institute (AISI) has published an initial evaluation of Anthropic’s Mythos Preview model, adding independent public verification to the company’s claims.
- While Mythos isn’t significantly different in tests of individual cybersecurity-related tasks compared to previous models, AISI highlights its ability to chain these tasks into multistep attacks necessary for system infiltration.
- AISI’s evaluation and subsequent Capture the Flag challenges show a steady improvement in AI model performance over time, with Anthropic’s Mythos Preview now capable of completing north of 85 percent of Apprentice-level CTF tasks. This underscores the ongoing need to carefully assess AI capabilities within cybersecurity contexts.
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.