“`html
- The UK government’s AI Security Institute has released an initial evaluation of Anthropic’s Mythos Preview model, providing independent public verification of its cybersecurity capabilities.
- This assessment shows that while Mythos isn’t significantly different in individual tasks from other recent models, it excels at chaining these tasks into multi-step attack sequences necessary to fully infiltrate systems.
- The evaluation comes as part of the UK government’s ongoing series of Capture the Flag challenges for AI models since early 2023, where performance has steadily improved with each new model introduced.
“`
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.