The recent launch of Anthropic's Claude Mythos model, specifically its first public iteration called Fable 5, has raised alarm among crypto users regarding its potential misuse. The company has implemented safety measures, but concerns remain that this AI model's capabilities could be exploited to compromise decentralized finance (DeFi) protocols.
Anthropic announced that Fable 5 identified over 10,000 high or critical-severity vulnerabilities in key software systems, leading to questions about the decision to publicly release it. The company reassured users that Fable 5 was "made safe for general use" and included safeguards that redirect sensitive topics, such as cybersecurity, to another model, Claude Opus 4.8. Still, the risks associated with releasing such a powerful AI tool have not gone unnoticed.
Simon Dedic, founder of Moonrock Capital, voiced his concerns on social media, pointing out that the ease and low cost of discovering exploitable flaws in smart contracts could increase dramatically. He warned that this development should serve as a major alert for DeFi projects, emphasizing that unaudited protocols could become easy targets for exploitation. Dedic stated, "Known exploits will get replayed on forks around the clock. Even small projects will get targeted simply because trying costs next to nothing now."
The crypto space has already felt the impact of AI-assisted hacks, with losses due to breaches reaching $629.7 million in April, the highest total since February 2025. As AI technology becomes more integrated into hacking tools, the implications for digital asset security are increasingly concerning. Dedic's calls for self-protection measures, such as revoking wallet approvals and transferring assets to secure hardware wallets, reflect the heightened vigilance among crypto users.
Not all industry experts, however, agree with Dedic's assessment of the threat posed by Claude Mythos. Michael Egorov, co-founder of Curve Finance, suggested that fears surrounding Mythos might be overstated. He noted that the software systems in which Mythos found vulnerabilities consist of millions of lines of code, in contrast to smart contracts that typically contain only a few thousand lines. Egorov remarked, "I suspect we might not be having a wave of DeFi code hacks, but we may see a lot of things in OpSec getting hacked (looking like multisig key compromises) and supply chain attacks on frontend dependencies."
Egorov's analysis suggests that while vulnerabilities exist, the nature and scale of potential exploits in smart contracts may not mirror those in larger systems targeted by Mythos. He believes that traditional and AI-assisted methods of vulnerability identification will not necessarily translate directly into the DeFi sphere.
Anthropic has also indicated that a select group of cybersecurity and infrastructure providers will gain access to the unguarded version of Claude Mythos, which may present additional risks, albeit under controlled conditions. This limited release highlights the need to balance leveraging advanced AI capabilities with ensuring robust security measures in the crypto ecosystem.
As AI and crypto continue to evolve, the implications of deploying powerful models like Claude Mythos will require ongoing scrutiny. Ongoing dialogue among industry experts and users will be crucial in navigating these challenges, ensuring that the benefits of AI do not compromise security in decentralized finance.
The stories that move AI & crypto markets — before the market reacts.
Free. 7am ET. Five stories. 62,400 readers.
