OpenAI's Joe warns safety-cyber divide risks 'great harm'

The post lands against a backdrop the article describes as full of contradictions. AI companies are releasing technology that enables sophisticated cyberattacks while also pitching the same technology as a defense: OpenAI has its Daybreak program and Anthropic has Project Glasswing, both granting select businesses access to advanced cybersecurity tools. Bad actors, meanwhile, are trying to use those same models — or open-source ones catching up quickly — for hacking. Joe's argument is that the two professional worlds are drifting apart even as the technology pulls them together.

The article notes Joe is not the first to flag this divide; former Fortune AI reporter Sharon Goldman has covered it too. It also acknowledges the post may irritate cybersecurity professionals, who could read it as calling them ignorant about AI — something the piece attributes to the AI industry's broader transparency problem, including the lack of standard disclosure frameworks for security incidents, which OpenAI is only starting to develop. Joe's personal framing — three months of "hell," skipping his sister's wedding — also drew criticism from those who noted he is building the very technology at issue.

Whether the suggestion changes anything likely hinges on whether the labs treat incident disclosure and joint decision-making as shared infrastructure rather than separate mandates. The test, for the safety and security teams named, is whether a seat at the table translates into practices both sides actually follow.