OpenAI Partners with Cathedral to Strengthen AI Security
As artificial intelligence keeps changing and getting better the dangers that come with cybersecurity are also becoming more advanced. People who try to break into systems are discovering methods to take advantage of problems in systems that use artificial intelligence. This makes checking the safety of these systems more necessary, than ever.
The collaboration wants to find spots, in OpenAIs systems and fix them before bad actors can take advantage. The company intends to use AI-driven security testing to make its digital infrastructure stronger. This will help its products better resist evolving cyber threats.
OpenAI joins hands with Cathedral Systems
OpenAI has confirmed that Cathedral Systems will participate in its upcoming internal security sprint. The initiative will focus on using artificial intelligence to identify potential weaknesses and improve the company’s existing security measures.
Cathedral Systems is a relatively new cybersecurity company founded by former DOGE staffers. The startup recently attracted attention after its AI agents reportedly identified security vulnerabilities in third-party business systems used by rival AI company Anthropic in less than 12 hours.
OpenAI hopes to benefit from Cathedral’s approach to automated security testing. By bringing external expertise into the process, the company wants to identify vulnerabilities that might otherwise go unnoticed during routine internal checks.
Bug bounty rewards increased to $172,000
Alongside the partnership, OpenAI is also making its bug bounty programme more attractive to independent security researchers and ethical hackers.
Under the updated programme, researchers who discover exceptional security vulnerabilities could receive rewards of up to $172,000, including bonuses. This represents more than four times the previous maximum payout for critical findings.
The increased rewards are intended to encourage experienced cybersecurity professionals to examine OpenAI’s systems and report serious security issues responsibly.
What happened during OpenAI’s previous security sprint?
OpenAI has already experimented with large-scale internal security exercises. During its previous sprint, the company reportedly assigned 25 per cent of its product engineers to identify and address security weaknesses.
Engineers used Codex to investigate reported issues, reproduce vulnerabilities in controlled testing environments and develop software patches to resolve them.
The latest initiative builds on those efforts, highlighting OpenAI’s intention to make security testing a regular part of its development process rather than treating it as a one-time exercise.
OpenAI explains the initiative on X
In an update shared on X (formerly Twitter), OpenAI highlighted how advances in AI could change cybersecurity. As AI models become increasingly capable of discovering vulnerabilities, companies must also use these technologies to protect their systems.
The collaboration with Cathedral Systems will let outside researchers work with the teams, at OpenAI. They will use the company’s models to test the security methods.
With partnerships testing that uses AI and much bigger rewards, for finding bugs OpenAI wants to stay ahead of more advanced online dangers. This effort also shows that more companies are trying to make sure that strong AI tools have security systems that’re just as good.