2026-11-03 · Source：MarkTechPost
Agent ID's Discover and Secure modules ship November 16, 2026, pairing MCP server visibility with per-call policy enforcement for finance, government, healthcare, and critical infrastructure.
One prompt did it. In the incident RSA's Jim Taylor invokes to frame the stakes, a single bad prompt took down a company's Salesforce — an autonomous agent carrying out an instruction no human had vetted.
The anecdote arrives with a larger number attached. RSA sizes the shadow problem at 4,000 AI agents running inside a single enterprise beyond the view of security teams, and it chose The AI Conference in San Francisco to launch Agent ID, an identity security platform built for agentic AI in finance, government, healthcare, and critical infrastructure.
The platform arrives as three modules with distinct jobs. Discover works as the reconnaissance layer, surfacing both sanctioned and shadow AI agents along with the MCP servers that connect agents to enterprise tools. Secure takes a different position in the stack: an inline gateway that vets each tool call against policy as it happens, rather than relying on a one-time approval granted at deployment. Govern, the third module, is pointed at the audit burden, mapping evidence to 10 regulatory frameworks.
For procurement planners, the timing is concrete: Discover and Secure ship on November 16, 2026.
The sequencing is telling. Visibility and real-time enforcement come before compliance reporting, on the logic that an agent nobody can see is an agent nobody can govern. The explicit coverage of MCP servers is equally deliberate — as the protocol hardens into a standard route from agents to tools, an MCP endpoint outside the inventory is by definition outside the control set. Whether per-call inspection scales cleanly across thousands of agents is the operational question buyers will test after the November ship date.
For banks, agencies, hospitals, and critical infrastructure operators, the pitch is blunt: the model is not the control point — identity is, and most agentic identities are currently invisible.
Tags: RSA · Agent ID · Jim Taylor · AI Agent Security · MCP Servers · AI Governance
萬安算交所 AIXX · AIXX · Verified compute exchange · Enter spot hall