AI Agent Security: Poor Credentials & Auditing Risks

Two factors can "corrupt" AI agent workflows, posing a dangerous combination for businesses. This is according to IBM senior solutions engineer Andre Faria and HashiCorp senior technical product marketing manager Van Phan. The first factor is giving AI agents long-lived static credentials that are poorly managed. The second is that these credentials are rarely rotated and tough to audit. This creates broad access with limited oversight. Without proper guardrails, AI agents could make decisions or take actions that negatively impact production, corrupt data, cause outages, or expose sensitive information. Organizations need to monitor active sessions, the systems AI agents access, and what actions they perform. They also need to check if agent behavior deviates from policy. This matters because AI agents require unique identities and just-in-time privileges for secure access.