AI is making vulnerability discovery and exploitation far cheaper in crypto, according to Guillemet, who said attackers are already using AI tools to uncover system weaknesses at a pace that was previously out of reach. Code flaws that once took seasoned security professionals months to find can now be identified in seconds. His warning was blunt: it has become extremely easy to identify and exploit flaws in a system.
Recent protocol exploits show the scale of the problem
Events from the past few weeks have reinforced that concern. This week, a major vulnerability in Drift, a DeFi protocol in the Solana ecosystem, was exploited for about $285 million in digital assets. A week earlier, yield protocol Resolv suffered an estimated $25 million loss. Data from DefiLlama puts the combined value of digital assets lost through cyberattacks or exposed vulnerabilities over the past year at $1.4 billion.
AI-assisted coding is also increasing exposure
Guillemet argued that secure systems should be costly and difficult to breach relative to the reward on offer. That balance is weakening. As developers lean more heavily on AI-assisted coding tools, new vulnerabilities can emerge faster and spread more widely across the crypto stack. He said there is no magic button that guarantees safety, and warned that the industry is likely to face a massive wave of code that is insecure by design.
Audit limits are becoming clearer
His view is that the crypto sector needs to rethink how it handles security at a foundational level. Traditional audits still matter, but they have limits, especially as protocols and supporting software grow more complex. He pointed to mathematical verification, or formally proving code correctness, as a stronger defense against vulnerabilities than standard audit methods in some cases.
Protection at the hardware level remains important as well. Hardware wallets can isolate private keys from internet access, reducing exposure to online attacks and adding a critical layer of defense.
Malware is shifting toward wallets and mobile devices
Threats are evolving at the user level too. Guillemet noted that malware now routinely targets wallet keys stored on compromised mobile devices, and in some instances assets can be drained without any user action. The pattern is clear: as AI and automation become easier to access, attack methods are becoming more sophisticated.
He urged users not to assume the systems they rely on are inherently safe, saying it is wiser to presume that most systems cannot be fully trusted. He also said practices such as cold storage, stronger authentication, and keeping sensitive data offline are likely to gain wider use among crypto holders, while software defenses alone will not be enough against both digital and physical threats.