As cybercriminals eagerly embrace AI tools, OpenAI needs all the help it can get securing its own systems from external hacks. That includes teaming up with a shadowy cybersecurity startup founded by a cadre of former DOGE employees.

The startup, called Cathedral, is best known for reportedly hacking into Anthropic’s private systems as part of the AI lab’s bug bounty program. Cathedral told the Washington Post that the hack was performed in under twelve hours and that it gained access to sensitive information within Anthropic, including Slack messages, job offers, and meeting transcripts. Anthropic cybersecurity lead Logan Graham responded in a X post that the hack actually took “several days,” and that the company’s “most sensitive systems, like model weights and Claude conversations,” were not affected. Graham also said Cathedral’s approach to the hack was “unusual and out of scope” of the rules Anthropic has in place for it’s bug bounty program, since it relied on multiple third-party vulnerabilities.

In a blog post published Monday, OpenAI said it was partnering with Cathedral as part of its broader effort to strengthen the security of its AI tools. Cathedral’s team “will use OpenAI’s models alongside our teams to challenge our defenses, bringing differentiated expertise and approaches to finding weaknesses and helping us protect our platforms,” OpenAI wrote. The company said it was also “more than quadrupling our rewards for critical findings” through its own bug bounty program. (Another company blog post from March of last year said the reward had quintupled from $20,000 to $100,000 This could be the “more than quadrupling” OpenAI is referring to, but the new post did not clarify.

Source link