Broadcom Integrates Identity Governance for Autonomous AI Agents

Broadcom Inc. is integrating identity-based governance into enterprise AI infrastructure to address security risks posed by autonomous agents. At VMware Explore 2026, the company outlined a strategy to apply distributed application tracing to AI prompts and tool calls, focusing on the core principles of identity, intervention, and inspection.

Clayton Donley, vice president and general manager of Broadcom's Identity Management Security Division, warned that businesses are deploying AI for mission-critical tasks without the established audit and certification regimes used for human employees. He noted that unlike the Sarbanes-Oxley era, where employee access required certification, no such standards currently exist for AI agents.

To close this governance gap, Donley recommends a passive initial approach where companies monitor traffic to identify agents. This process leads to the implementation of a central control point designed to enforce policy and block the use of unauthorized API keys from providers such as OpenAI or Claude.