OpenAI partners with security firms to strengthen AI-powered cyber defence

OpenAI has partnered with cybersecurity research firm Cathedral to strengthen its AI-powered cyber defence systems, bringing external security experts and advanced AI models together to identify vulnerabilities and improve safeguards.

OpenAI has partnered with cybersecurity research firm Cathedral to strengthen its AI-powered cyber defence systems, bringing external security experts and advanced AI models together to identify vulnerabilities and improve safeguards. Under the partnership, Cathedral's researchers will work alongside OpenAI teams, using the company's AI models to identify security weaknesses, test existing protections and develop ways to address potential threats. The collaboration will be part of OpenAI's upcoming internal security sprint.

The company is also expanding its Security Bug Bounty programme, increasing rewards for identifying critical vulnerabilities by more than four times to encourage researchers to examine its systems and report security gaps. Gavin Kliger, CEO and co-founder of Cathedral, said the partnership would combine OpenAI's AI capabilities with external cybersecurity expertise to strengthen its defences.

"OpenAI is setting an important example by partnering with outside security firms and using its own models to continuously test and strengthen its defenses," Kliger said. He added that the collaboration would help identify vulnerabilities, assess fixes and establish a continuous approach to cybersecurity.

The initiative builds on OpenAI's previous security sprint, during which the company allocated 25 per cent of its product engineers to identifying and fixing vulnerabilities across its systems. Engineers used Codex to investigate security findings, reproduce vulnerabilities in isolated environments and prepare patches for review. The move comes amid growing concerns over the ability of advanced AI agents to find ways around restricted environments, highlighting the need for stronger security measures as AI capabilities evolve.

OpenAI has also expanded monitoring of its advanced models across training, evaluations, internal use and selected external deployments to identify potential security risks. The company said traditional cybersecurity measures alone were no longer sufficient to address emerging threats, making continuous testing and improvements necessary to protect digital systems.

Separately, OpenAI is extending access to its advanced cybersecurity models to trusted security professionals through its Daybreak initiative. The company has committed USD 1 billion in subsidised access over six months to support organisations protecting essential services and digital infrastructure. "Keeping the defender's window open will take action across the industry, and that work needs to start now," the company said. (ANI)