Lawfare AI Liability Symposium

I'm participating in this in-person event today at the Georgetown Law Center in D.C., and Lawfare is posting the articles here (the link also includes some other pieces on related topics). The articles all look very interesting; mine in particular is on AI and the First Amendment—readers of the blog might find it familiar, since it's based on past articles (including one cowritten with Mark Lemley and Peter Henderson), but it might still be a helpful shortish digest of those longer articles.

Comments

[Roger S]

In libel cases, the threshold “key inquiry is whether the challenged expression, however labeled by defendant, would reasonably appear to state or imply assertions of objective fact.” ... What, for instance, is OpenAI doing when it promotes ChatGPT’s ability to get high scores on bar exams or the SAT?

It is not claiming that any particular assertion is a fact. It is merely claiming that it scores well on some tests.

Likewise when AI software is incorporated into search engines, or into other applications, presumably precisely because it’s seen as pretty reliable.

Yes, but so far the search engines have escaped liability. If I get this blog from searching "crooked law professors", the search engine does not have any liability.

[Davy C]

No, you get this blog if you search for "crook scorpion lawyer". Because...

In general, when you search for such things, you expect websites talking *about* the crooks, not the website *of* the crooks. (You also wouldn't be all that surprised if it showed you pages about a lawyer who once had a case involving scorpions, or a comic book villain scorpion named Hook Crook, or a law firm from Crook County.)

But now, with AI, you face the possibility of having a blurb *above* the search results stating that Professor Roger Scorpion was convicted of bribery last week. Would a reasonable person think that such a blurb is accurate? If it's never accurate then why even show it?