TLDR Tech / InfoSec
UK Gov's Mythos AI Tests Help Separate Cybersecurity Threat From Hype
Anthropic's Mythos Preview model faced UK AI Security Institute tests on capture-the-flag tasks and a 32-step “The Last Ones” network data exfiltration range, where it became the first AI system to complete the full chain in some runs and averaged 22 steps versus Claude 4.6's 16, while still failing a tougher “Cooling Tower” power-plant scenario and operating in ranges without active defenders or realistic detection penalties.
Read the full story