CertiK CEO: AI Agent Access Risks "Looming Disaster

The CEO of CertiK, Ronghui Gu, warns that giving AI agents access to sensitive systems like local files and financial infrastructure is a "looming disaster." What's happening is autonomous AI systems are being deployed more and more in enterprise tools and financial applications. These AI agents can manage workflows, analyze data, and even execute transactions. To do this effectively, they need deep access to user environments, including personal data and banking APIs. Gu highlights that this widespread access creates a large attack surface. This surface could be exploited through vulnerabilities, malicious prompts, or compromised AI behavior. The concern is that combining autonomous decision-making with unrestricted system access could lead to unintended consequences, especially with financial transactions or confidential data. This matters because the increasing use of AI agents could introduce significant new security risks for everyone.