←
Danny Wilf-Townsend1,329 posts
Danny Wilf-Townsend avatar

Danny Wilf-Townsend

@drmtown

Working on artificial intelligence, consumer protection, and civil procedure. Law professor currently visiting at @YaleLawSch, usually at @GeorgetownLaw

◷ Joined July 2009▧ 136 Photos and videos
Danny Wilf-Townsend avatar
Danny Wilf-Townsend@drmtown·32m

This seems overstated to me. First, these lawsuits could easily fail; just filing a lawsuit under an old law doesn't prove much. Second, even if broad old laws can get there on the substance, statutes can put in processes that are much faster and more reliable than litigation.

For all the hubbub over new AI safety laws, it’s really telling that this lawsuit invokes nuisance and unfairness laws that date to the mid-19th Century. All of those tremendously expensive, laborious fights over California’s first-in-nation frontier AI safety law. To what end?

Danny Wilf-Townsend avatar
Danny Wilf-Townsend@drmtown·20h

So, at least in my book, this suit has moved into the number one spot for cases to watch re public safety and AI. /end.

Danny Wilf-Townsend avatar
Danny Wilf-Townsend@drmtown·20h

...the litigation will take. There are lots of opportunities for delay. But especially with the arguments about existing investigations being thwarted, I could imagine the AG getting some form of preliminary relief in this case, especially around information or transparency. 8/

Danny Wilf-Townsend avatar
Danny Wilf-Townsend@drmtown·20h

There will still be challenges, of course. These are new territories for consumer protection and public nuisance law, raising questions of how well the harms fit existing doctrine. And, as always, there is the question of how long ... 7/

Danny Wilf-Townsend avatar
Danny Wilf-Townsend@drmtown·20h

It's doable for state AGs to make policy through court injunctions/settlements. It's less common to do this at the scale that some litigants have sought for frontier AI companies. But it's not clear from the complaint just what the AG will seek, and it may be feasible. 6/

Danny Wilf-Townsend avatar
Danny Wilf-Townsend@drmtown·20h

The AG also has an advantage when it comes to a challenge I pointed out yesterday, i.e. that it's a lot to ask a court to take on a broad AI regulatory project. This is still a big ask, but the AG brings credibility, legitimacy, and resources to bear. 5/

And that gets me to the third hurdle here. These cases are on some level asking an individual judge to take over the project of national AI regulation. The legal theories, especially nuisance, are extremely open-ended. That's asking a lot of a judge. 8/

Danny Wilf-Townsend avatar
Danny Wilf-Townsend@drmtown·20h

So California has a clearer argument for being a strong plaintiff here—a closer connection to injury, and more direct regulatory responsibility, than many other plaintiffs. And an existing failure to comply with investigation is an easier sell than an uncertain future harm. 4/

Danny Wilf-Townsend avatar
Danny Wilf-Townsend@drmtown·20h

The AG also has authority to investigate OpenAI, and an important throughline in this complaint is OpenAI's failure to respond adequately to the AG's investigation. This is an easy thing for a court to identify as a concrete, legitimate problem. 3/

Post attachment
Danny Wilf-Townsend avatar
Danny Wilf-Townsend@drmtown·20h

This may be obvious, but you know who the California AG represents? Californians. This includes California businesses, an extremely relevant constituency for harms related to hacking, shared internet infrastructure, and resources like RubyGems and HuggingFace, etc. 2/

Danny Wilf-Townsend avatar
Danny Wilf-Townsend@drmtown·20h

When these other AI public safety suits have been filed, one of my reactions each time has been "this feels like an argument that would be more effective coming from the California AG." And now the California AG has filed. A few thoughts on why this suit is different. /1

Post attachment

Between this, the Florida AG suit, and the LASST suit, there is now a small wave of lawsuits challenging frontier AI companies on the basis of broad public safety concerns. They involve different causes of action, and address real risks, but I think they share similar hurdles. 1/

Danny Wilf-Townsend avatar
Danny Wilf-Townsend@drmtown·Nov 16

But in general, the courts move too slowly and are typically too focused on individual, ex post injuries to be the path that I would bet on for public safety regulation of the AI industry. /end

Danny Wilf-Townsend avatar
Danny Wilf-Townsend@drmtown·Nov 16

I could be wrong about all of this, of course. There are some very good lawyers and organizations working on these suits. And in general, it's good to have a regulatory movement have a court strategy as part of it. So I'll be watching these suits and others like it. 13/

Danny Wilf-Townsend avatar
Danny Wilf-Townsend@drmtown·Nov 16

If the plaintiffs win a battle over a preliminary injunction, that would change things, but I would not bet on that. And without preliminary relief, the timeframes for litigation make it hard to see it playing a major role in AI regulation around public safety. 12/

Danny Wilf-Townsend avatar
Danny Wilf-Townsend@drmtown·Nov 16

And finally, the last hurdle is timing. Even if the judge is willing to move through these challenges, the defendants have more than enough tools to make these cases take years, especially if the direction looks bad for the defendants. I'd think three, four years, no problem. 11/

Danny Wilf-Townsend avatar
Danny Wilf-Townsend@drmtown·Nov 16

Judges are going to be wary of the scope of enjoining a major industry in a highly technical domain on an extremely open-ended theory of liability like nuisance. The degree of supervision would be intense, the number of regulatory judgment calls would be vast. 10/

Danny Wilf-Townsend avatar
Danny Wilf-Townsend@drmtown·Nov 16

It is true that we need much better AI regulation, and the courts in theory could serve as a forcing function for that. And courts have shown in the past how the law can adapt to meet new needs for regulating industries. So I'm sympathetic to that. But... 9/