Brockman pushes OpenAI into cybersecurity, sparking safety clash
TL;DR
- OpenAI president Greg Brockman tells The Information he is in 'full management mode', steering the company toward corporate customers and cybersecurity tools.
- OpenAI has reassigned about 25% of its production engineers to defend its own systems and committed $1 billion to frontline defenders.
- The cybersecurity push has coincided with OpenAI rewriting its Preparedness Framework, the governance document that decides when a model is too dangerous to ship.
OpenAI president Greg Brockman told The Information he is in 'full management mode' as he steers the company toward corporate customers and cybersecurity, telling reporter Jemima McEvoy there is no question cyber will be a bigger business than coding.
'I have a feeling that cyber spend, security spend should actually vastly outstrip software spend,' Brockman has argued, and the receipts he cites are internal. OpenAI took 'about 25% of our production engineers' off their projects and pointed the company's most capable model at its own infrastructure, committed $1 billion to frontline defenders, and says roughly $1 million of compute was enough to identify about 1,000 Linux vulnerabilities and produce patches for them.
The commercial push has opened a parallel fight over safety. The day after Brockman's 'Defender's Window' essay, OpenAI said it is rewriting its Preparedness Framework, the internal governance document that sets when a model is too dangerous to develop further. Reporting around the interview frames the move as a showdown over AI safety rather than a routine refresh.
The cyber momentum shows up elsewhere on our tracker: NATO's NCIA now runs up to 300 billion OpenAI tokens a month through Daybreak, the same enterprise program Brockman is pushing to the center of the company's roadmap.
Originally reported by theinformation.com
Read the original article →Original headline: OpenAI's Greg Brockman Tells The Information He's in 'Full Management Mode' on Cybersecurity