The Impact of Anthropic's Mythos Model on Crypto Industry Security

The introduction of Anthropic's Mythos AI model has sparked a significant shift in the crypto industry's approach to security. For years, the focus has been on securing smart contracts through auditing and vulnerability testing. However, Mythos, with its ability to identify and exploit weaknesses across entire systems, is driving attention towards the infrastructure that supports these contracts. According to Paul Vijender, head of security at Gauntlet, a risk management firm, 'the bigger risks sit in infrastructure.' He emphasized that when considering AI-driven threats, the focus should be on AI-assisted attacks against human and infrastructure layers, rather than just smart contract exploits. This includes key management systems, signing services, bridges, oracle networks, and the cryptographic layers that connect them. These components are often less visible and outside the traditional audit scope, making them more vulnerable to attacks. A recent security breach at web infrastructure provider Vercel, which many crypto companies use, exposed customer API keys, prompting crypto projects to rotate credentials and review their code. The breach was attributed to a compromised Google Workspace connection via a third-party AI tool. Mythos belongs to a new class of AI systems designed to simulate adversaries, exploring how protocols interact and testing how small weaknesses can be combined into real-world exploits. This approach has drawn attention beyond the crypto space, with banks like JP Morgan treating AI-driven cyber risk as systemic and exploring tools like Mythos for stress testing. Early findings from models like Mythos have identified weaknesses in the behind-the-scenes systems that keep crypto platforms secure, including the technology that protects keys and handles communication between systems. Vijender noted that AI models are especially valuable in identifying multi-step exploit chains and infrastructure-layer vulnerabilities that traditional audits may miss. The shift towards AI-driven security matters in a system built on composability, where DeFi protocols can connect and build on each other's services. This interconnectedness has driven growth but also creates pathways for risk to spread, as seen in recent bridge exploits. Without AI, those dependencies are hard to trace, but with AI, they can be mapped and exploited at scale, resulting in a shift from isolated exploits to systemic failures that cascade across protocols. Some industry leaders see Mythos as an acceleration rather than a turning point. Stani Kulechov, founder of Aave Labs, believes AI reflects the dynamics already at play in DeFi's adversarial environment. 'Web3 is no stranger to well-funded and motivated adversaries,' he said. 'AI models represent an evolution in the tools used to achieve exploits.' Kulechov noted that DeFi is already built for machine-speed attacks, and AI intensifies an environment that has always required constant vigilance. Aave is seeing AI surface new categories of vulnerabilities, including issues that human auditors may have previously deprioritized. To defend against offensive AI, the answer lies in changing the security model itself. Audits before deployment and monitoring after were designed for human-paced threats, but AI compresses that timeline. 'To defend against offensive AI, we will need to take an AI-centric approach where speed and continuous adaptation are essential,' Vijender said. Aave has already integrated AI into its workflows, using it for simulations and code review alongside human auditors. 'We take an AI-first approach where it adds clear value,' Kulechov said. 'But it complements, rather than replaces, human-led auditing.' For builders, the long-term effect may be less disruption than divergence. Hayden Adams, founder and CEO of Uniswap Labs, expects the gap between secure and insecure protocols to widen over time. 'Projects that prioritize security will have greater ability to test and harden systems before launching,' he said. 'Projects that don’t will be most at risk.' Ultimately, security is no longer about eliminating vulnerabilities but about continuously adapting to a system in which those vulnerabilities are constantly rediscovered and recombined.