Cloud-based HR and finance giant Workday announced Agent Passport this week at its DevCon conference. Every agent that touches HR, finance, or IT data inside Workday now needs to pass a test before going live, and gets closely watched after.

What makes this more than a checkbox feature is who’s doing the checking. Cisco is the launch partner, using its AI Defense product to independently test agents against prompt injection, jailbreaks, goal hijacking, system prompt extraction, and data leakage. Each test ties back to a named public standard, OWASP’s LLM Top 10, NIST’s AI Risk Management Framework, MITRE ATLAS, and gets signed by whoever ran it.

That independence is the actual pitch. Most agent security testing today is done by the same vendor that built the agent, marking its own homework. Workday’s bet is that a signed, third-party attestation means something a self-issued one doesn’t, especially once you’re trying to compare an agent built in-house against one from a third-party vendor on the same platform.

Once an agent is live, Agent Passport keeps watching. Every action gets allowed, blocked, or routed based on policy. Find a problem and a single revocation can shut the agent down, or restrict it, across the business in one move.

For a platform sitting on payroll, benefits, and employee data for more than 11,500 organisations, that kind of kill switch isn’t a nice-to-have. It’s the bar enterprise buyers will start expecting from anyone selling agentic tools into HR and finance.

Early access lands in the second half of 2026, with general availability targeted before year end. Whether other enterprise platforms follow with their own independently-attested passport system, or push back on Workday setting the standard, is the thing worth watching next.