Anthropic launched Claude Fable 5 today, June 10, a public version of its previous version Mythos. Now, the most powerful AI risk detection built in is only available to everyone who signed up. For DeFi, the results are immediate.
Until today, Mythos was locked behind Project Glasswing, owned by about 150 hand-picked organizations, including Google, Microsoft, and JPMorgan. The brand had already found more than 10,000 problems in the world’s most important software.
What Legend 5 Can Do And What You Can’t Do
Anthropic says Legend 5 comes with a safe margin. In high-risk areas, including cybersecurity, biology, chemistry, and sample distillation, the model restricts responses and reverts to Claude Opus 4.8.
Anthropic depression its testers who tried to jailbreak before the release of Fable 5, are running foreign currency that did not result in a jailbreak after more than 1,000 hours of testing.
Cybersecurity requests trigger a refund less than 5% of episodes, meaning that great interactions continue throughout Fable 5’s action.
For DeFi in particular, the use of smart contracts does not fit well in the restricted Anthropic environment. Finding a vulnerability in Solidity’s contract looks more like a scripting exercise than a typical cyber security attack.
The unique performance of Fable 5 in programming is flexible, and the longer and more complex the task, the more advanced it is than other models available. For an immutable DeFi protocol running on publicly available code, this difference can be very important.
The Zcash model is already circulating in security circles. A light version of Anthropic architecture he found a big problem in the Zcash protocol within 24 hours, an attack that survived four years from some of the most famous experts in the world.
Why DeFi Is Specially Revealed
Legend 5 alone represents the change in value and skill needed to search smart contracts for vulnerabilities. White hat hacker MevenRekthe explained the situation clearly: the cost and skill required to find possible errors in the smart industry is almost down to zero.
Countless protocols are sitting ducks. Common features can be repeated on forks throughout the season. Even the smallest projects are worth looking at because of the money-for-money test.
Anthropic itself warned last week that AI systems are advancing so quickly that they will soon be able to improve themselves repeatedly, improving without human intervention.
The advice from security experts is solid and quick: change token approvals, transfer funds to hardware wallets, and limit exposure to protocols you don’t trust. Not tomorrow. Today.
A note Claude Fable 5: The World’s Most Powerful Destructive AI Should Have DeFi Worried appeared for the first time BeInCrypto.
