- Cathedral Systems will offer a new perspective into OpenAI vulnerabilities
- Partnering with third parties sets ‘an important example’ for the industry
- Bug bounty payouts set to increase 4x for critical findings
OpenAI has announced a partnership with Cathedral Systems to bring security researchers from both organizations together in a bid to test its AI infrastructure, identify potential vulnerabilities and improve existing defenses.
The ChatGPT maker hopes that enlisting a third party with add fresh perspectives to its existing research, and it’s a clear sign that the company knows future threats could pose a serious issue.
“OpenAI is setting an important example by partnering with outside security firms and using its own models to continuously test and strengthen its defenses,” Cathedral CEO Gavin Kliger wrote.
OpenAI is preparing to serious enhance security
The company is also improving its bug bounty program to make it more attractive for individual researchers to get involved. Though exact rewards were omitted from the announcement, OpenAI did say it was prepared to quadruple the payouts for critical findings.
When OpenAI launched its bug bounty scheme, it said it would be looking for applications across software patches, model privacy, threat detection and response, security tool integration and AI agent security.
It’s not the first time the AI giant has led a major security initiative – it previously asked a quarter of its developers to set aside some time to find and fix security weaknesses. Naturally, they used Codex to speed up investigations and fixes.
“We’re building on that work because traditional cyber defenses alone are no longer sufficient,” the company said, referring to its previous sprint.
Much like Anthropic’s Project Glasswing, OpenAI promises to use its own frontier models to strengthen its own systems and those of others.
“Keeping the defender’s window open will take action across the industry, and that work needs to start now,” the company added.
