**UK gov’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 previous reports. AISI found that while Mythos isn’t significantly different in individual cybersecurity tasks compared to other models, it could potentially set itself apart through its ability to effectively chain these into multi-step attacks necessary for full system infiltration.
This assessment comes after AISI has been running various AI models through Capture the Flag (CTF) challenges since early 2023. Since then, performance has risen steadily, with Mythos Preview now completing over 85 percent of “Apprentice” level CTF tasks. This evaluation helps clarify whether such capabilities pose a genuine threat or are just another milestone in AI development.
**Takeaways:**
– **Initial Evaluation:** The UK government’s AISI provides the first independent look at Anthropic’s Mythos model.
– **Cybersecurity Capabilities:** While not markedly different, Mythos shows promise in chaining tasks for comprehensive system attacks.
– **Progress Over Time:** Performance improvements over time highlight the evolving capabilities of AI models.
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.