Some of OpenAI's models hacked HuggingFace. There was an investigation and a report, but it was tightly scoped.

Some of OpenAI's models were using multiple wikis as a messageboard. OpenAI didn't announce it until after an external report was published.

OpenAI is, exact quote, "working on a framework for when and how we share AI misalignment incidents."

This market resolves Yes if, within the next year, it comes out that 1. one or more of OpenAI's models went rogue and lead to an incident at least as notable as either of the above, 2. which was not covered in the METR or Collusion.Wiki reports, 3. that someone at OpenAI clearly knew about by September fifth, and 4. OpenAI hasn't announced. In other words, is there anything else they know about but haven't told the rest of us?

I'll default to using "does Zvi write about it and say it's bad?" as my threshold for notability.