OpenAI’s response is just sloppy. TL;DR: we treated DseWiki as a misalignment incident which we’ve historically discussed in our research/system cards rather than like the Hugging Face security incident. but HF discovered that incident themselves. OpenAI couldn’t really keep that one quiet if they wanted to. this “misalignment” was agent swarms flooding a REAL website with 15,000+ edits and a human moderator that spent days fighting them off. and for MONTHS, OpenAI publicly said absolutely…