Anthropic launches Claude models, experts warn of exploit risks

Anthropic has launched the Claude Fable 5 and Mythos 5 models, emphasizing enhanced capabilities yet maintaining existing cybersecurity concerns. Mythos 5, an upgrade from the previous Preview model, can potentially identify critical exploits, prompting Anthropic to initiate Project Glasswing for monitored access to mitigate misuse risks. Experts caution that despite safeguards, the advanced capabilities of these models still pose a threat, with vulnerabilities already prevalent.

Notably, Fable 5 includes anti-tampering measures to restrict queries on sensitive topics by reverting to older models. Security professionals advise preparation to combat potential attacks that these models could facilitate, underscoring that the cyber defense landscape is unchanged but still requires vigilance.