chevron-down Created with Sketch Beta.

ARTICLE

AI Strikes Again: If You Can’t Trust Your Own Expert, Who Can You Trust?

Neil B Posner

Summary

  • Artificial intelligence sometimes creates fictitious citations.
  • Counsel’s reliance on the work of experts must be reasonable under the circumstances.
  • There can be no substitute for verifying that any citations actually exist and verifying that they support your case.
AI Strikes Again: If You Can’t Trust Your Own Expert, Who Can You Trust?
d3sign via Getty Images

In 2023, Minnesota enacted a new statute, Section 609.771, titled “Use of Deep Fake Technology to Influence an Election.” Under that statute, a person who “disseminates a deep fake or enters into a contract or other agreement to disseminate a deep fake is guilty of a crime . . .  if the person knows or reasonably should know that the item being disseminated is a deep fake . . . .”

Two individuals—Christopher Kohls, a social-media personality, and Minnesota State Rep. Mary Franson—brought a suit against the State of Minnesota in Federal Court, seeking a preliminary injunction blocking the state’s ban on the dissemination of election-related AI-generated content.

The Minnesota attorney general, in opposing the suit, obtained an expert opinion from Professor Jeff Hancock, co-director of Stanford University’s Cyber Policy Center and director of the center’s Social Media Laboratory. In his declaration, Hancock said that the Minnesota attorney general retained Hancock to provide expert opinion and testimony “regarding how AI is influencing misinformation on social media and the psychological impact of deceptive deep fakes, particularly deep fakes shared on social media.” Hancock further stated in his declaration that he is a “leading expert in social media behavior and the psychology of online interaction.”

According to Kohls and Franson, however, Hancock’s declaration included a citation to a study that “bears the hallmarks of being an artificial intelligence ‘hallucination,’ suggesting that at least the citation was generated by a large language model like ChatGPT.” As it turns out, the study Hancock cited—purportedly in the Journal of Information Technology & Politics—does not exist. That journal is real, but the article is not.

In light of plaintiffs’ discovery that the expert’s declaration contained a hallucination, they filed a Daubert motion to exclude expert declarations. See Kohls & Franson v. Ellison, Case No. 24-cv-03754-LMP-DLM, Doc. 30 (D. Minn., Nov. 16, 2024). In their motion, they stated that if the lawyers inserted the fictional citation in the declaration, plaintiffs intend to pursue sanctions. But if Hancock or his assistants generated the content, that is a more serious matter because Hancock has stated that he has been an expert witness in twelve different matters over the last five years.

If Hancock originated the hallucination, this court might order him to inform the parties and courts in any pending litigation about the possibility [that] his testimony may contain content generated by a large language model. The court has inherent authority to prevent fraud by the parties appearing before it.

Motion at p.9.

***

The applicable ABA Model Rule of Professional Conduct—1.1 (Competence)—continues to control. Comment [8] to Rule 1.1 provides:

To maintain the requisite knowledge and skill, a lawyer should keep abreast of changes in the law and its practice, including the benefits and risks associated with relevant technology, engage in continuing study and education and comply with all continuing legal education requirements to which the lawyer is subject.

(Emphasis added.)

While it is certainly understandable why counsel would seek out the opinions and testimony of experts in a particular area, including (or especially) emerging technology, counsel’s reliance on the work of such experts must be reasonable under the circumstances. The Terminology Rules of the Model Rules (Rule 1.0) provide three definitions that refer to the concept of reasonableness:

(h) “Reasonable” or “reasonably” when used in relation to conduct by a lawyer denotes the conduct of a reasonably prudent and competent lawyer.
(i) “Reasonable belief” or “reasonably believes” when used in reference to a lawyer denotes that the lawyer believes the matter in question and that the circumstances are such that the belief is reasonable.
(j) “Reasonably should know” when used in reference to a lawyer denotes that a lawyer of reasonable prudence and competence would ascertain the matter in question.

(Emphasis added.)

And, as we can see from this unusual but cautionary case, counsel is not relieved of her responsibility to verify everything that goes into a motion to a court. By “verify,” we mean not only that the citations therein actually exist but, further, that such citations actually support the arguments being made to the court.

Accordingly, regardless of whether the citations are the result or product of your (or your assistant’s) research or the product of your expert, there can be no substitute for verifying that the citations actually exist and verifying that they support your case.

    Author