“HuggingFace like” == publicly known misalignment incident that occurred contrary to the wishes of a frontier lab during training, internal experimentation, or some other similar scenario. Leaks or rumors won’t count, this has to be something publicly acknowledged by the lab, with enough publicly released details to assess the magnitude.

The incident must of equal or larger scale than the HF-OpenAI one, which would likely involve a ton of discussion on Twitter/ACX-adjacent blogosphere at the very least. The scale would be judged by how much discussion it generates in relevant communities vs by absolute “scale”.

Question will resolve based off the spirit rather than the letter of the market. It’s a “you know it happened when you see it” kind of market.

• Update 2026-09-12 (PST) (AI summary of creator comment): - News reporting and discussion occurring prior to 2026-08-30 will not be taken into account.

  • The underlying incident itself could have occurred at any time in the past.

• Update 2026-09-25 (PST) (AI summary of creator comment): - Incidents that are erroneously perceived as misalignment but are later confirmed not to be will not resolve to YES, even if they generate widespread discussion or scandal.