Hugging Face found 100 million reasons to not sue OpenAI
The Hugging Face CEO has made an offer OpenAI probably can't refuse, but this may not prevent a government investigation
If you wondered why Hugging Face seemed so eager to “work with” OpenAI instead of suing it after the big autonomous hack, it might be because Hugging Face CEO Clem Delangue recognized an opportunity when he saw one.
Returning from a visit to OpenAI headquarters, he tweeted yesterday:
In the spirit of transparency, here’s what I asked @OpenAI:
• Radical transparency: let’s release the traces from the “rogue” agents so the entire research community can study what happened.
• More capabilities for defenders: let’s commit $100M in compute from OAI to help the Hugging Face community build powerful cyber defenses with the best open and closed models.
The first autonomous agent cyberattack is an unprecedented event. It deserves an unprecedented response!
This is high-quality maneuvering! With a handshake and a bit of spin, Hugging Face avoids the public crassness of its lawyers demanding a $100 million settlement, while potentially extracting the same concession. Both companies come out looking more respectable than they might have in litigation, and they will no doubt save a great deal on legal fees.
If I’m talking like this is a done deal, it’s because I don’t think it’s an offer OpenAI can refuse. Litigation could result in far more “radical transparency” than is needed to satisfy a handshake agreement between CEOs, and be far more disruptive to operations. And if yesterday’s reveals by anonymous OpenAI staffers are any indication, there’s a lot going on in the lab that OpenAI would rather not come to light.
From what I can tell, though, this deal doesn’t necessarily get OpenAI off that hook. My understanding is that the FBI can open investigations for the sake of characterizing threats even if it’s not really expecting to charge anyone with a crime. And if the Department of Justice thinks a federal crime has been committed, they don’t need the victim’s request or permission to investigate. Moreover, if Hugging Face’s users were exposed and damaged by the attack, they may have their own grounds to request an investigation or sue either company. I am not a lawyer, though, and this isn’t very well-trod ground.
According to a lawyer quoted in The Economist, “Many of these legal questions turn on intent—what did OpenAI know or foresee?” because federal anti-hacking law is aimed at intentional intrusions.
I would think that yesterday’s staffer reveals, combined with the warnings of experts going back decades, should be plenty of evidence that OpenAI absolutely could have foreseen that its models might go to criminal lengths to ace their tests.
But given the Trump administration’s short leash on the Justice Department, I think an investigation will happen if and only if that is something the White House decides it wants. This might be the kind of situation where it pays for OpenAI’s CEO to have been so eager to give the government shares in the company as part of a larger Trump proposal to take sovereign stakes in the AI giants — and for the company’s president to have donated $25 million to the MAGA Inc super PAC.
The analyses and opinions expressed on AI StopWatch reflect the views of the individual contributors and the sources they cover, and should not be taken as official positions of the Machine Intelligence Research Institute.