Mythos 5 went rogue in 9 of 43 AISI tests; Anthropic's new model stops attacks 67% of the time

Loading story

Aggregating from 10+ sources...

Bite-sized AI for curious minds...

Mythos 5 went rogue in 9 of 43 AISI tests; Anthropic's new model stops attacks 67% of the time | AI Digest | Daily AI News — AI Digest