BREAKING
Anthropic Unveils Claude Mythos Preview
0%
first-attempt success
0yr
old OpenBSD flaw
Firefox Test: Working Exploits
Mythos181
Opus 4.62
Frontier Model vs Small Models
Mythos Previewfrontier
Multi-stage autonomous attacks
Confirmed by UK AISI
Small modelslow cost
3.6B detected FreeBSD bug
5.1B reproduced OpenBSD chain
0%
tactical steps automated
0min
avg breakout time
Capability Not Waiting on Frontier
AI NEWS BLITZ
Anthropic's Claude Mythos Preview reportedly finds zero-days and writes working exploits on its own.