Tanium joins OpenAI cyber programme for threat research

Tanium has joined OpenAI's Trusted Access for Cyber programme, giving it access to OpenAI's cyber models for threat research and platform security.

It is using the programme in three areas: threat research and detection; vulnerability and remediation research; and security work related to its software, cloud services and infrastructure.

Tanium's threat hunters are using AI reasoning tools to study attacker techniques and emerging threats. That research feeds into detection content distributed to customers.

Engineers are also using OpenAI GPT Cyber Models to explore new ways to investigate vulnerabilities, prioritise remediation and address weaknesses. Tanium said this work is shaping how it helps customers reduce risk.

At the same time, its security engineers are incorporating the models into existing internal security practices to defend the company's platform against new attack methods.

Cyber focus

The arrangement comes as cybersecurity groups face a shift in how attacks are planned and carried out. Tanium said attackers are already using AI to identify weaknesses, blend into normal activity and move faster at greater scale.

For defenders, the same technology offers a way to apply automation and reasoning to tasks that have often required significant specialist time and expertise. Access through the programme is limited and not broadly available.

Tanium added that its use of the programme includes human oversight and must operate within OpenAI's authorised-use boundaries. These requirements include identity verification, access governance and permitted-use safeguards.

Christian Hunt, Chief of Engineering at Tanium, described the development as part of the company's broader security work.

"Customers trust Tanium with visibility and control across their entire environment, and we take that responsibility seriously," said Christian Hunt, Chief of Engineering at Tanium. "TAC puts OpenAI GPT Cyber Models in the hands of our threat hunters and security engineers. What they learn informs stronger detections, new approaches to reduce risk, and a platform customers can depend on as attackers increasingly use AI."

Tanium is known for software used by corporate and public sector technology teams to monitor and manage endpoints across large estates. The company said the research and security work tied to the programme reaches customers through the platform they use to protect critical endpoints.

AI pressure

The announcement highlights a broader push by cybersecurity suppliers to incorporate generative and reasoning-based AI into defensive operations and internal product security. The pressure has increased as threat actors experiment with AI-assisted reconnaissance, evasion and social engineering.

Vulnerability management is one area where suppliers see scope for practical use. Sorting large volumes of flaws, deciding which issues present the greatest risk and matching fixes to operational constraints has long been labour-intensive, especially in large organisations with complex endpoint environments.

Threat research is another field where AI tools are being tested in practice, particularly for analysing attacker behaviour, mapping techniques and supporting detection engineering. Tanium said findings from that work are translated into the detection content it already provides to customers.

The company framed its participation as an extension of its existing security practices rather than a standalone product launch. It said the models are being applied to the research and engineering work behind the platform, rather than being presented as a separate customer-facing service.

Tanium also tied the move to a broader rise in AI-related risk across IT and security operations. The company, which describes itself as focused on autonomous IT, said organisations need to keep pace with threats increasingly shaped by attackers' use of AI.

Hunt said Tanium's work under the programme will centre on detection, risk reduction and platform defence as the threat landscape changes.