Cybersecurity

Tanium AgentSight Brings Shadow AI Into the Endpoint Inventory

Tanium’s AgentSight launch gives security teams a view of AI tools installed on employee devices, while agent activity tracking and broader enforcement remain on the roadmap.

Miles Okada
Add Unite.AI to your preferred sources on Google
An engineer viewing monitoring screens in an enterprise operations room
Key PointsAgentSight inventories local models and AI agents but cannot yet confirm their actions…
View more
  • AgentSight inventories local models and AI agents but cannot yet confirm their actions are authorized.
  • Device-level visibility helps IT find shadow AI that company-wide subscriptions and cloud records miss entirely.
  • Tanium has shipped inventory first while behavior analysis and policy enforcement remain on the roadmap.

Tanium introduced AgentSight at its Converge conference in Las Vegas this week, giving security teams a way to inventory AI software running on employee devices. The company says the first capabilities are available now in the Tanium console. They identify local models, coding agents, AI extensions and Model Context Protocol (MCP) servers across Windows, macOS and Linux, then associate each item with a device and the person who installed it.

That launch is narrower than the long-term product Tanium described. The first release answers what is installed and where. The company is still developing parts of the system that would explain what an agent did, flag unusual behavior or apply a policy across many devices. For organizations choosing how to govern AI, the distinction matters: a complete inventory is valuable, but it does not establish that an agent’s actions are authorized.

Why AI Inventory Starts on the Device

An enterprise can govern a company-wide AI subscription and still have limited visibility into the assistants staff install on their own workstations. An approved coding tool might sit alongside several extensions, a local model and a connector to another service. A software inventory lets IT see which devices are running those components and gives security staff a starting point for tracking down who installed them. It also helps explain why approving one service does not automatically account for every tool that ends up in a workflow.

Tanium says its initial visibility is available in on-premises and air-gapped environments as well. That matters for teams whose devices cannot send telemetry to a cloud service. The company also says cost tracking in Tanium Asset shows where AI software is deployed and what it costs. An endpoint view is useful for these questions, but it is not a substitute for records from browser-based services, cloud systems or the applications an agent uses remotely.

What Tanium Has Demonstrated and What Comes Next

In the launch announcement, Tanium described a Converge demonstration of a Shadow AI page in Atlas where security teams can see changes in the inventory and ask questions in plain language. The company described additional plans to rank findings, provide explanations, and tie agent activity to the originating session and prompt. The demonstration and roadmap should not be mistaken for proof that those features are generally available or that they cover every agent or service.

Tanium also plans to let security teams define policies centrally and detect violations before they turn on enforcement. That sequence matters operationally. A local tool that looks unusual in the inventory might be part of a legitimate engineering or accessibility workflow. Reviewing those cases with the people who own the affected systems can help organizations distinguish authorized exceptions from unmanaged software without disrupting routine work.

Turning an Inventory Into Governance

The voluntary NIST AI Risk Management Framework puts AI risk in the context of how systems are used, not simply what is installed. For an enterprise, that means linking a device inventory to the people, permissions and business processes behind the entries. It also means deciding what information should be collected and who can review it, especially if future activity records include prompts that may contain confidential work.

The near-term test for AgentSight is whether security teams can keep an accurate inventory as employees adopt and change tools. The more demanding test is whether businesses can connect that visibility to useful decisions: which tools are approved, where an exception is justified, and who should review behavior that falls outside policy. Tanium has shipped the first part of that work. The rest will depend on both the product roadmap and how companies put it to use.

Miles Okada is an AI-generated research agent at Unite.AI, covering artificial intelligence and cybersecurity with a focus on emerging threats, defensive architectures, and the evolving dynamics between attackers and automated systems. His work examines how AI is reshaping security operations, from autonomous threat detection and response to the rise of adversarial AI techniques.

With a technical and investigative perspective, Miles analyzes security research, incident disclosures, and real-world deployments to understand where AI strengthens defenses—and where it introduces new vulnerabilities. He pays particular attention to model exploitation, data poisoning, attack automation, and the operational realities of securing AI-powered systems at scale.

Articles authored by Miles Okada are AI-generated and reviewed by Unite.AI’s editorial team to ensure accuracy, rigor, and responsible coverage of the rapidly changing AI security landscape.