Anthropic's most powerful AI model raises DeFi hack risks, experts warn

Anthropic's Claude Fable 5 and the restricted Mythos 5 model can find and chain software vulnerabilities at superhuman speed. DeFi, which has lost over $840 million to hacks this year, is one of the most exposed industries if safety filters fail.

• Anthropic released Claude Fable 5 publicly and a more powerful Mythos 5 variant only for vetted cybersecurity users
• Experts say the AI won't invent new hack types but will dramatically speed up finding misconfigurations and exploiting human errors
• DeFi has already lost more than $840 million in hacks this year, mostly from operational failures not smart-contract bugs
• Ledger's CTO said current AI guardrails 'are not a reliable control against a determined adversary'
• Anthropic admits it expects well-funded attackers to keep trying to bypass safety measures

Coinporta summarises news for fast scanning. Full reporting belongs to the source.