Claude Fable 5 Could Turn Hours into Minutes for Crypto Hacks

Anthropic released Claude Fable 5 on Tuesday, its most powerful public reasoning model, alongside a restricted version called Claude Mythos 5 for cybersecurity teams. The company says Mythos can find and chain zero-day vulnerabilities – software flaws no one has seen before – and turn a bug into a working exploit. For crypto markets, the timing is stark.

DeFi protocols lost more than $840 million to hacks in the first five months of this year, DefiLlama data shows. April alone accounted for $600 million, the worst month on record. The two largest incidents – a $285 million drain from Drift Protocol and a $292 million theft from Kelp DAO – did not rely on smart-contract exploits. They used social engineering and key compromise. A third hit on Tuesday: Humanity Protocol, a decentralized identity service, lost over $30 million after a hacker gained access to three of six private keys on one employee's laptop.

Charles Guillemet, chief technology officer at hardware-wallet maker Ledger, said AI's main contribution is not inventing new attack types but accelerating the reconnaissance phase. “A reasoning model can diff every commit, grep every config, and enumerate every misconfiguration at machine speed,” he said in an email. The core vulnerabilities remain familiar: weak signing flows, exposed private keys, human error.

Anthropic acknowledges the system is not foolproof. In a blog post, the company said fewer than 5% of session requests trigger a fallback to a weaker model, Claude Opus 4.8. Even after more than 1,000 hours of bug-bounty testing, researchers found no universal way to bypass the guardrails. The company expects well-funded attackers to keep trying because the payoff is large. “The uplift from Mythos-level capabilities is valuable to many adversaries – for instance, those who could financially gain from cyberattacks – and we therefore expect them to be motivated to try to circumvent our safety measures,” the firm said.