SecurityBrief UK - Technology news for CISOs & cybersecurity decision-makers
United Kingdom
Tanium launches AgentSight to track shadow AI on devices

Tanium launches AgentSight to track shadow AI on devices

Wed, 18th Nov 2026 (Today)
Sofiah Nichole Salivio
SOFIAH NICHOLE SALIVIO News Editor

Tanium has launched AgentSight, a product designed to identify and manage artificial intelligence tools running across corporate endpoints. The release is aimed at organisations dealing with unauthorised or poorly understood AI use on employee devices.

It is intended to give IT and security teams visibility into AI software installed across Windows, macOS and Linux systems, including local models, coding agents, MCP servers, desktop AI applications, integrated development environment extensions and browser add-ons. Each detected component is mapped to the device where it is running and to the person who installed it.

The launch comes as businesses grapple with broader concerns over so-called shadow AI, where employees adopt AI tools without formal approval from IT departments. In material accompanying the announcement, Tanium said companies are increasingly dealing with software that can act autonomously while using employee permissions and access rights.

This creates a challenge for organisations trying to understand which AI tools are active across their estates, who is using them and whether they create security or compliance risks. AgentSight is intended to extend Tanium's existing endpoint monitoring to AI-related software and services.

According to Tanium, the product provides visibility through the Tanium console, including in on-premises and air-gapped environments. It also integrates with Tanium Asset to show what AI software is costing an organisation and where it has been deployed.

One highlighted feature is a shadow AI page in Tanium Atlas, which presents an inventory of detected AI tools and tracks changes from the previous day. Teams can query the system in natural language and ask for explanations of specific findings.

Risk monitoring is another focus. Tanium said the system is intended to rank AI-related risk findings, explain the reason behind each one and maintain a record of actions taken by an agent, the order in which they occurred and the prompt that initiated them.

"An agent misusing its access may look like legitimate activity, so you need to know what's expected and normal. Tanium already builds behavioural baselines for endpoints, and AgentSight extends that discipline to agents," said Harman Kaur, Chief Technology Officer at Tanium.

"It will rank AI risk findings and give a plain reason for each. It will also record what each agent did, in order, to the session and the prompt that started it, and alert when an agent departs from its own normal. Endpoints, agents, the infrastructure they depend on, and the identities behind them are managed as one estate on one platform," Kaur said.

Some response measures are already available through Tanium's Threat Response product, including killing processes, quarantining devices and using Live Response tools. The company is also developing broader policy controls that would allow security teams to set AI rules once and take action on devices running software that falls outside those policies.

Any wider blocking or removal functions would begin in a detect-only mode. Tanium said this is intended to reduce the risk of disrupting legitimate business tools before IT teams have assessed their role and impact.

AgentSight enters a market where employers are trying to balance employee demand for AI assistants with concerns over data exposure, unauthorised software procurement and the difficulty of tracing actions taken by autonomous tools. Businesses have faced similar issues with shadow IT for years, but AI introduces extra uncertainty because software can generate outputs, trigger workflows and interact with corporate systems in less predictable ways.

The release also comes against a backdrop of growing public policy concern in the UK over cyber resilience and AI-related threats. Tanium referenced warnings from the House of Lords that the country is underprepared for emerging cyber risks, including those linked to artificial intelligence.

AgentSight sits as a separate product alongside Tanium's existing Security Operations, Endpoint Management, Exposure Management and Atlas offerings. Visibility functions are available through the Tanium console from today, while additional control features remain in development.

Some response capabilities, including kill process, quarantine and Live Response, are available today through Threat Response.