- METR has identified that its current test suite struggles to adequately measure Claude Mythos, with only five out of 228 tasks covering relevant capabilities.
- Palo Alto Networks warns of the growing threat posed by autonomous AI attackers, noting that these models can now autonomously exploit vulnerabilities and exfiltrate data within just 25 minutes from initial access.
- There is a significant gap in evaluation methods for newer AI models like Claude Mythos, which may be more vulnerable to such attacks due to their rapid evolution compared to established tools.
Originally published at the-decoder.com. Curated by AI Maestro.
Stay ahead of AI. Get the most important stories delivered to your inbox — no spam, no noise.