They absolutely could. There's a decent claim for OAI being reckless or at least negligent.

The limiting thing, in my mind, is the need to show harm. How much did OAI's recklessness hurt HuggingFace? There's some dev hours devoted to the cleanup but it's not like it devastated their business. Their legal team is probably a lot more focused on the Nvidia acquisition instead.

I feel like it's the government that has to sue OpenAI for reckless endangerment of human life, or something similar. I don't believe the current models could "take over the world" even if humans tried to help them, but both OpenAI and Anthropic themselves talk publicly about how dangerous they believe these models to be. Then if that's so, they should be held responsible as if the models were that dangerous, but who's supposed to put that in motion? The only entity I can think of is the government.

Would there also not be some reputational damages that could be compensated for as well? These are the types of cases usually ending in a settlement for non-disclosed amounts with NDAs attached.