A single prompt causes salesforce outage: rsa's jim taylor discusses agent ID and the hidden threat of shadow AI agents

At The AI Conference in San Francisco, RSA introduced Agent ID, an agentic identity security platform for regulated industries. Jim Taylor, RSA’s President and Chief Product and Strategy Officer, warned of AI agents’ risks, including accumulating permissions and acting autonomously. A customer success employee at an unnamed company inadvertently triggered a Salesforce denial-of-service attack by requesting an agent to download all data, leading to the instance being shut down. RSA’s Agent ID includes three modules: Discover, Secure, and Govern. Discover scans endpoints and applications in real time, identifying sanctioned and shadow agents. Secure enforces policy checks on tool calls, while Govern logs actions and maps them to regulatory frameworks. Taylor emphasized the need for human oversight and proper ownership of agents, warning against unchecked AI behavior. RSA plans to release Discover and Secure by November 16, 2026, with Govern following in early 2027. Taylor also addressed the growing threat of shadow AI agents, estimating over 4,000 such agents in a single global bank. He stressed integrating agentic security with existing measures, noting CI/CD pipelines as an identity attack surface. He argued agents should not inherit permissions from others, closing privilege-escalation paths. For CISOs evaluating Agent ID, Taylor recommended starting with a pilot by connecting key systems and running a Discovery scan, which often reveals unexpected agent activity.