August 19 Update: Law360 reports NetChoice will drop this expert, and provides a response from NetChoice. I’ve added a small addendum at the end of this post.
There’s a troubling filing over in the United States District Court Middle District Of Louisiana, where a group dedicated to free expression is alleged to have allowed its own expert to be a little too free.
The case is NetChoice vs. Liz Murrill et al., Civil Action No. 3:25-cv-231. By way of background, Liz Murrill is the Attorney General of Louisiana, and the case at hand is a challenge to Louisiana’s SB162 law, which requires social media companies to verify the ages of their users. This is a wholly interesting area of law in its own right, but today we’re popping in just to highlight a particularly problematic turn the litigation has taken.
According to the Defendants’ August 14 filing, “Defendants’ Memorandum In Support Of Their Motion To Exclude Dr. Anthony Bean”, NetChoice had identified Dr. Anthony Bean, a Fort Worth psychologist, as its sole expert. NetChoice served Dr. Bean’s declaration, the report in question, titled “Expert Review of the Alleged Causal Relationship Between Social Media Use and Youth Mental Health Outcomes” on the Defendants. But the Defendants’ rebuttal expert quickly found “herself chasing ghosts, fabricated quotations from articles that do not exist” in preparing a response.
Defendants now allege that the “substance” of the report submitted by the Plaintiff’s expert, Dr. Bean, is “fake.” In particular, they allege “Dr. Bean’s report bears all the telltale signs of AI hallucinations.” Altogether, Defendants allege “None of the 17 articles in Dr. Bean’s reference list exists…none of the 12 quotations that Dr. Bean’s report attributes to various authors and articles exists.”
These accusations are embarassing enough for Dr. Bean and NetChoice, but the Defendants dig in just a bit:
In fact, just reading Dr. Bean’s report would have done so. His reference list makes no sense, (a) citing website links that are dead or lead to entirely unrelated sources and (b) citing volume and page numbers in publications that are easily confirmed to be wrong. And his report itself is strangely formatted, not least because, well, it looks and reads like a print-out from artificial intelligence (AI).
Defendants ask that Dr. Bean (and his report) be excluded, that NetChoice not be allowed to replace him as an expert, and for reasonable costs and fees.
What Went Wrong?
There’s a lot to unpack here. If these allegations are true, there are at least three questions that jump off the page at me…
First of all, shouldn’t NetChoice or its attorneys be competent enough in this subject matter to spot an AI-generated report? Probably. Someone should have been able to tell the details in this didn’t pass the smell test.
Second, shouldn’t NetChoice or its attorneys be competent enough to spot AI when they see it? Maybe, maybe not. AI gets continually more sophisticated. And where writing is somewhat standardized, it can be trickier to spot a fraud.
Third, shouldn’t NetChoice’s attorneys have done basic due diligence in reviewing this report? Yes. As it turns out, experts lied and falsified data and report before AI was around. Some of them are now a bit lazier and using easy-to-spot AI, but the real issue here isn’t AI, it’s the failure of the attorneys to check the report they submitted.
One paradox of this moment of the AI age is that it’s maybe easier than in the recent past to spot shoddy work product. When it comes to blogging, we’re past the era when one can easily spot a post written by some hired freelancer with no experience. There are enough good ghostwriters out there to write content that passes the sniff test. AI-generated content is, for now, pretty easily identifiable. To that end, people who are lazy enough to just call on an AI to write for them are going to be quickly spotted.
Addendum: In their comment to Law360, NetChoice seems to be saying AI was not involved and this was a more traditional tech-related oopsie. This would address my first two comments on “what went wrong”, leaving the third—someone needed to do their job a little better before hitting the submit button.
