OpenAI is partnering with security firm Cathedral Systems to bring researchers from both organizations together to test the company’s AI infrastructure, find potential vulnerabilities and strengthen its defenses. The move is intended to add outside perspectives to OpenAI’s security work.
The company also plans to increase the payouts in its bug bounty program. OpenAI did not disclose specific reward amounts, but said payments for critical findings will increase fourfold.
OpenAI’s bounty program seeks reports covering areas including software patches, model privacy, threat detection and response, security tool integration and AI agent security.
The partnership follows an earlier security effort in which OpenAI asked a quarter of its developers to spend time finding and fixing weaknesses, using Codex to support investigations and repairs. The company said it is expanding that work because “traditional cyber defenses alone are no longer sufficient.”
Cathedral Systems CEO Gavin Kliger said working with outside security firms and using OpenAI’s models to test and improve defenses sets an important example for the industry.
Sources: TechRadar.