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

The introduction of Mythos, Anthropic's novel AI model, has sparked widespread concern and confusion within traditional tech and finance, while simultaneously driving a significant shift in the crypto industry's approach to security. For years, the primary focus of decentralized finance has been on defending smart contracts through auditing, vulnerability cataloging, and understanding common exploits. However, Mythos, designed to identify and chain together weaknesses across systems, is now pushing the industry to look beyond code and into the underlying infrastructure. According to Paul Vijender, head of security at Gauntlet, a risk management firm, 'The greater risks lie in the infrastructure.' He emphasized that when considering AI-driven threats, his primary concern is not smart contract exploits, but rather AI-assisted attacks targeting human and infrastructure layers. These components include key management systems, signing services, bridges, oracle networks, and the cryptographic layers connecting them. Since these elements are less visible than smart contracts and often fall outside traditional audit scopes, they pose significant risks. Recently, web infrastructure provider Vercel, which is widely used by crypto companies, disclosed a security breach that may have exposed customer API keys. This incident prompted crypto projects to rotate credentials and review their code. The breach was attributed to a compromised Google Workspace connection via the third-party AI tool Context.ai, used by an employee. Mythos represents 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 industry, 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 secure crypto platforms, including technologies protecting keys and handling inter-system communication. Vijender noted, 'I think AI models are particularly valuable in two areas: first, multi-step exploit chains that are typically only discovered after funds have been lost, and second, infrastructure-layer vulnerabilities that traditional audits often overlook.' This shift is crucial in a system built on composability, where DeFi protocols can interconnect and build upon each other's services. The interconnectedness of DeFi protocols has driven growth but also creates pathways for risk to spread, as seen in recent bridge exploits. 'Composability is what makes DeFi capital-efficient and innovative,' Vijender said. 'However, it also means that a minor vulnerability in one protocol can become a critical exploit vector with contagion potential across the ecosystem.' Without AI, tracing these dependencies is challenging. With AI, they can be mapped and exploited at scale, resulting in a shift from isolated exploits to systemic failures that cascade across protocols. The evolution of AI attacks has led some industry leaders to view Mythos as an acceleration rather than a turning point. Stani Kulechov, founder of Aave Labs, believes AI reflects the existing dynamics 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.' From this perspective, DeFi is already built for machine-speed attacks, with smart contracts executing automatically and defenses operating without human intervention. 'AI doesn't introduce a new dynamic; it intensifies an environment that has always required constant vigilance,' Kulechov said. Despite this, Aave is discovering new categories of vulnerabilities using AI, including issues that human auditors may have previously deprioritized. The breadth of AI-driven threats still matters in a system where even smaller vulnerabilities can undermine trust or be combined into larger exploits. If attackers can move faster, the question becomes whether defenses can keep pace. For both Gauntlet and Aave, the answer lies in changing the security model itself. Audits and monitoring 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. This includes continuous auditing, real-time simulation, and systems built with the assumption that breaches will occur. Aave has 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.' In this sense, AI equips both attackers and defenders. For builders, the long-term effect may be less disruption than divergence. Hayden Adams, founder and CEO of Uniswap Labs, expressed interest in what Mythos and similar tools can do for protocol security. 'AI gives builders better ways to stress test and harden systems,' he said. Over time, Adams expects the gap between secure and insecure protocols to widen. 'Projects that prioritize security will have a greater ability to test and harden systems before launching,' he said. 'Projects that don't will be most at risk.' This may be the real shift. Security is no longer about eliminating vulnerabilities; it's about continuously adapting to a system where those vulnerabilities are constantly rediscovered and recombined.