The agent security gap: 54% of enterprises have already had an AI agent incident, and most still let agents share credentials

A concerning trend emerges from a study on AI agent security, revealing that over half of the 107 surveyed enterprises have already experienced a security incident involving their AI agents. Despite this, many organizations still do not enforce proper credential management, allowing agents to share credentials and only a third provide unique, scoped identities to each agent. This lax approach to security could lead to significant vulnerabilities, especially given that most enterprises rely on existing security frameworks rather than developing specialized controls for AI agents. The implications are severe, as it underscores a critical gap in enterprise security strategies that could be exploited by malicious actors.