OpenAI hired hackers who had previously hacked Anthropic in less than 12 hours. The company announced a partnership with the startup Cathedral, whose experts will look for vulnerabilities in OpenAI products using its own neural networks. In the past, Cathedral staff, as part of the bounty program, were given access to Anthropic's internal systems, including slack conversations and meeting transcripts. OpenAI also plans to more than quadruple the rewards for critical vulnerabilities.