OpenAI to bring Cathedral into a new security review and raise vulnerability rewards
Prepared by Sintelligent with the help of SI. Published in SI: . Updated: .
The same models that can find coding errors are becoming a permanent defense tool for SI platforms, while external review receives a stronger incentive.
According to OpenAI's description, Cathedral researchers will work with OpenAI teams and use the company's models to find weaknesses and test fixes. In the previous internal sprint, OpenAI assigned 25% of its product engineers to finding and fixing vulnerabilities. Rewards for critical findings in core products have more than quadrupled: the base maximum is $30,000, exceptional findings can earn up to $100,000, and quality, loyalty, and special-campaign bonuses can bring the total to approximately $172,000. Researchers who reported vulnerabilities in the previous 30 days will receive the difference from the old rates retroactively.
Sources 2