OpenAI fires workers for 'mishandling sensitive information'

There’s a delicious irony lurking in this story. OpenAI—an organization founded on the promise of open, democratic AI—has now terminated workers for "mishandling sensitive information." The information in question wasn't sold to foreign spies or posted on a dark web forum; it was shared with an outside AI evaluation group. That's right: a group whose entire purpose is likely to assess safety, alignment, or performance. In other words, the people trying to hold AI accountable were punished for reaching out to other people trying to hold AI accountable. This speaks to a deeper tension at the heart of the AI industry. We want our AI systems to be transparent, auditable, and safe—but we also want the companies building them to remain competitive, secretive, and profitable. When those two desires collide, secrecy almost always wins. The firing of these employees signals that OpenAI views internal information as proprietary, even when that information might be critical for external safety evaluations. It's a strange form of regulatory capture: the regulators are the evaluated, and the evaluated are the regulators. From a broader perspective, this is about the sociology of AI safety. The field has long struggled with the "information hazard" problem: knowing too much about a powerful system might enable misuse, but knowing too little prevents oversight. OpenAI's decision to fire employees suggests they believe the former risk outweighs the latter. But one has to ask: if an AI evaluation group is deemed untrustworthy, what does that say about the trustworthiness of the AI itself? And what does it say about the trust we should place in a company that treats its own employees as potential leakers? We may never know the full details of what was shared, or why. But this incident is a chilling case study in the emerging power dynamics of the AI era. As machines grow smarter, the humans who build them are being asked to hold their tongues. The question is whether that silence will protect us—or just protect the people in charge. Source: [BBC News - OpenAI fires workers for 'mishandling sensitive information'](https://www.bbc.co.uk/news/articles/c6y9z9r4ejzwo)