- Anthropic publicly released Claude Fable 5, its newest Mythos-class AI model.
- The model is the first to top 90% on its internal benchmark for long analytical tasks.
- Experts warn that Fable 5 could make it easier to find vulnerabilities in smart contracts.
A new AI model from Anthropic is potentially raising alarms in the crypto world, especially among DeFi developers and security experts.
Anthropic unveiled Claude Fable 5 to the public, a version of its much‑anticipated Mythos‑class AI that had been limited to a handful of governments and cybersecurity researchers.
The company says Claude Fable 5 is its most powerful public model yet. It reportedly hit state-of-the-art marks in software engineering, complex reasoning, analytics, and code analysis. According to Anthropic, the model is also the first to top 90% on its internal benchmark for long analytical tasks.
Fable 5 shares the same core architecture as Claude Mythos 5, which is a tightly restricted model that turned heads earlier this year for its remarkable skill at spotting software bugs. Anthropic initially kept Mythos under wraps, worried its cybersecurity powers could be abused by threat actors.
The company added safeguards that block certain queries involving cybersecurity, biology, and chemistry. If users try something deemed as high risk, the system defaults to an older, weaker model, Claude Opus 4.8. Anthropic says it thoroughly tested these protections before release.
DeFi May Be Uniquely Exposed
While the company has built safeguards into the public release, experts warn that the model could make it much easier to find vulnerabilities in smart contracts and blockchain apps.
For instance, Moonrock Capital’s founder and managing partner, Simon Dedic, discussed the situation with a white hat hacker, MevenRekt. Both talked about how the model drastically lowers the barrier for finding and exploiting vulnerabilities in smart contracts and protocols.
They believe the time and expertise needed to spot exploitable bugs in smart contracts is about to fall to basically zero.
As such, worry in the crypto world isn’t that Fable 5 will show people how to hack protocols outright. The problem lies in the fact that smart contract security sits in a gray zone.
Unlike hacking a bank or similar systems, reviewing public smart contract code is usually seen as legitimate analysis. Since DeFi protocols put their code out in the open, advanced AI could find flaws much faster than human auditors.
Until now, Mythos has been kept behind Project Glasswing, available only to about 150 select groups, including Microsoft, JP Morgan, and Google. It had already uncovered over 10,000 critical vulnerabilities in the highly popular software.
Related: SBI Holdings Partners With Anthropic for Group-Wide Claude AI Rollout
Disclaimer: The information presented in this article is for informational and educational purposes only. The article does not constitute financial advice or advice of any kind. Coin Edition is not responsible for any losses incurred as a result of the utilization of content, products, or services mentioned. Readers are advised to exercise caution before taking any action related to the company.
