> The facts are pretty bad.

The facts alleged in initial complaints usually do look pretty bad. But the causes of action which are still standing after summary judgement and discovery tend to look a lot less bad. So I tend to heavily discount, if not outright dismiss, the facts alleged in initial complaints. But I'm jaded having seen too many damning alleged facts evaporate before trial.

To be clear, I'm no fan of OAI or Altman. I think Altman, Brockman et al did pretty much subvert a non-profit. Unfortunately, I also think the OAI's board (past and recent) screwed up in a series of remarkably bone-headed ways which allowed Altman, Brockman, et al to get away with it. It's ironic that some of OAI's board members most aligned with the EA (Effective Altruism) movement ended up being so ineffective in fulfilling the governance duties of a non-profit board member.

But the fact OAI is run by dudes who could have been sent by central casting for the role of "Smarmy Asshole Tech Bro #1 & #2" nor that ChatGPT said unimaginably horrific things to vulnerable people, fulfills ALL the prongs of each test required to get a conviction on these causes of action. Even if ChatGPT literally gave Hitler a detailed plan and an upbeat "Okay, let's do this!" to executing the actual Holocaust, that's still just the first prong. And hitting the first prong a million times harder doesn't move any of the other required prongs even a bit.

Even in cases where the accused is obviously guilty, proving all the required prongs beyond a reasonable doubt can still be very hard. It's going to come down to a bunch of the details I mentioned in my other reply to you like: sales and usage context as well as claimed features of the product along with disclaimers, disclosures, existing practice, prior knowledge of actual harm, average user competence, etc. In short, it's not enough that the 'product' actually caused horrific things. IANAL and I don't know Florida liability law but, in broad strokes, cases like this need to prove that an average reasonable person would clearly know these exact bad things were likely to happen based on what OAI knew at the time, and that OAI had an express duty to prevent those things, and was directly negligent in fulfilling those duties and knew they were negligent. Each of those involve nuanced situational judgements based on a lot of facts and timing not yet in evidence and if just one of them isn't quite provable beyond a reasonable doubt enough to convince 100% of a Florida jury, the whole case can fail. LLM chatbots being so new and unknown creates a lot of reasonable doubt for the prosecution to overcome. That's why I'm confident the prosecution has no desire to actually try this case.