This post will be short (fingers crossed) and
sweet bitter. I’ve written pretty extensively lately about the two lawyers (now 3 lawyers) getting a lot of negative publicity for misusing ChatGPT and, as a result, filing documents with courts containing imaginary legal authorities. I will not repeat my points about the failings of those lawyers.
Today, instead, I will criticize three (so far) federal judges for having an entirely disproportionate, mostly unhelpful, and potentially problematic overreaction to AI issues. Almost immediately after the news broke about the first two lawyers to get called on the carpet for treating ChatGPT as a search engine, a federal judge in Texas issued a long administrative order that included a new certificate that would have to be added to every court filing by every lawyer in that court. A little while thereafter a federal judge in Illinois issued a similar administrative order and also required a certificate disclosing the use of generative AI.
I had kind of hoped that those two events would not be the beginning of a cascade, but today I learned that a federal judge in Pennsylvania has also now imposed (as of about 5 days ago) a requirement of a new certification to be added to every court filing in that court.
I will say three things.
Thing 1: These kinds of certifications are not necessary to address the failings of lawyers. The ethics rules already require lawyers not to make false statements of the law to courts and the rules of civil procedure already require that when a lawyer signs a court filing, they are certifying certain things to be true. This is not that different than if, after the first time some lawyer foolishly showed up to a court hearing drunk, courts started making lawyers swear before starting to argue that they hadn’t had anything to drink.
Thing 2: All that requiring yet one more certification to be made to court filings will do is further (even if only just incrementally) drive up the expense of litigation. Someone filing something in federal court already has to include form language to certify that they served the document on certain people, and if it is a motion (for the most part), they have to certify that there was an appropriate pre-filing consultation using certain standard language, and if it is a brief (for the most part) they have to certify that they complied with rules related to word counts and font-sizes and the like. Even if all this does is waste attorney time to add yet another certificate, it still just wastes client’s money paying for that time.
Thing 3: Further, using this new certification as an example, using AI is not the problem, it is misusing AI that is the problem. While the Texas requirement is still overkill, it at least attempts to be clear about tackling mostly the important issue. Likewise, Illinois (and frankly even better than Texas because it focuses more clearly on generative AI usage rather than general usage and reminds lawyers that Rule 11 already covers the landscape.) Given the wording of the Pennsylvania requirement, it is not at all clear that every lawyer who used spell-check on a filing would not have to disclose that fact. And, perhaps even more to the point, given how pervasive AI is behind the scenes in many different types of programs, it would be hard for an attorney to say anything truthfully in response to this certification requirement other than: “I’m not entirely sure whether I used AI in the process of preparing this filing, but I can certify that I’ve checked all of my cites and authorities personally before filing.”
Which, if that is what will be produced in the end as a result of this requirement, see Thing 1 and Thing 2 above.