AI Models Bring New Risks and Opportunities
This week, the intersection of artificial intelligence and decentralized finance security took a significant step forward. Anthropic introduced Fable 5, its first publicly released Mythos-class AI model, featuring a built-in safety-routing mechanism. When users submit queries related to cyberattacks or other sensitive topics, the system automatically redirects them to a less advanced model, Claude Opus 4.8. This approach reflects growing concerns that advanced AI could accelerate the discovery of vulnerabilities in privacy coins, smart contracts, and DeFi protocols, making both attack and defense more complex.
The debut of Fable 5 is part of a broader trend: as AI capabilities grow, their impact on DeFi security becomes more pronounced. Both defenders and attackers now have access to increasingly sophisticated tools, which has made the security environment more unpredictable. The same AI systems that help identify critical flaws are also being used to exploit them, raising the stakes for all participants in the ecosystem.
AI Empowers Both Attackers and Defenders
Recent findings show that attackers are already using AI-assisted tools to target DeFi protocols. In the past six months, at least $36.7 million has been stolen from protocols relying on unverified smart contracts. Attackers are leveraging AI to analyze raw bytecode and uncover vulnerabilities at speeds previously unattainable. This development challenges the traditional belief that code obscurity can offer meaningful protection, as AI can now rapidly bypass such barriers.
On the defensive side, AI is also proving its value. Earlier Anthropic models, including Claude Opus 4.8, have played a key role in uncovering vulnerabilities, such as a critical flaw in Zcash’s Orchard protocol. Automated AI auditing is helping teams identify subtle bugs before they can be exploited. However, the effectiveness of these tools depends on their integration into broader security practices. Without ongoing review and a commitment to act on AI-generated findings, even advanced solutions may not be enough to prevent attacks.
How DeFi Teams Should Respond
For DeFi projects, the message is clear: relying on unverified or obfuscated smart contracts is increasingly risky as AI tools make it easier for attackers to find and exploit weaknesses. Security must become a core part of protocol design and daily operations, rather than an afterthought. This requires adopting transparent development practices, integrating automated auditing, and fostering a culture of continuous risk assessment and improvement.
Protocols that adapt quickly by combining human expertise with AI-driven security measures are better positioned to protect user funds and maintain trust. The rapid pace of AI development means that security strategies must also evolve quickly to keep up with emerging threats and new attack methods.
What to Watch in the Weeks Ahead
The relationship between AI and DeFi security is set to deepen further. As new AI models are released, they will likely continue to expose vulnerabilities at a faster rate, and attackers may become even more adept at using these tools. The DeFi community should expect to see more protocols adopting automated auditing, opening their codebases, and rethinking risk management frameworks to address these ongoing challenges.
For more on how attackers are exploiting unverified smart contracts, see the latest research from Chainalysis: Attackers Exploiting Unverified Smart Contracts.
To compare secure onchain routes and find more efficient options, visit the Chainspot Router.
