Ping Identity has launched what it calls “Enterprise Personal Agent Access”, a capability aimed at helping organisations identify and govern personal AI agents, including desktop assistants such as Claude and coding agents such as Claude Code.
The company said the offering is designed to provide discovery of AI agents in use, associate agent sessions with the user and device behind them, and apply runtime controls over what agents can access and do. Ping said the capability is delivered through its PingOne Privilege product and is available now, with pilots underway with global enterprises.
The announcement comes amid growing concern about “shadow AI” and the security implications of autonomous agents being used in enterprise environments. Ping cited the Gravitee State of AI Agent Security report, which it said found 48% of production AI agents are running unsecured.
Unlike human users, agents can execute tasks across multiple systems at machine speed and scale, raising questions about access governance and post-incident accountability. The company argues that traditional identity controls built primarily for humans are not sufficient for this shift.
“Enterprises are deploying a number of different AI models, agents and platforms,” said Andre Durand, Founder and CEO of Ping Identity. “The security challenge is establishing a common way to see and govern all of them. The question isn’t whether an agent is intelligent enough to act, but whether the enterprise can see and control its action when it does.”
Ping said the technology is intended to support scenarios where personal AI agents interact with enterprise resources beyond the AI application itself, including MCP servers, code repositories, internal services and APIs, Kubernetes clusters, databases and cloud infrastructure. In such environments, the company said enforcement at the “point of action” is required to determine whether actions should be allowed, denied, logged, require human approval, or have access revoked in real time.
For developer workflows, Ping said the approach enables agents to reach approved resources and commit code without holding long-lived credentials, and that activity can be attributed to the agent as well as associated with a user session.
Ping also said it participated in Anthropic’s Project Glasswing, which it described as evaluating advanced AI capabilities to strengthen the security and resilience of its codebase and internal systems.
