Four vendors gave AI agents their own identity layer

BigID launched AgentIQ on September 22. It runs a data security and compliance program from a prompt inside Claude, Copilot, GPT or Gemini, through BigID's own MCP connector. Three other vendors released products the same day making an identical bet.

A static service account doesn't get revoked when one agent misbehaves. It doesn't get restored after a bad run touches the wrong table either. These four products assume that gap now costs a budget line.

Cohesity introduced Agent Resilience, built to discover, protect and recover the infrastructure behind enterprise AI agents, starting with support for Amazon Bedrock. Semarchy shipped Intelligent Data Products, where every data product automatically spins up its own MCP server carrying its quality rules and access rights. SereneDB released a database that enforces roles and permissions for agents at the data layer itself.

These four companies compete in different markets: security, backup, master data management and databases, respectively. But they all reached the same answer to one question. Does an agent get an identity of its own, or does it run on someone else's login?

None of the four vendors say what happens when an agent's permissions conflict across two of these systems. BigID's MCP layer and SereneDB's role enforcement could both sit on the same database with no shared rulebook. That gap is where the next incident starts.