No ChatGPT in court: judge orders AI content checked

Using ChatGPT to petition or present defenses in Judge Brantley Starr’s court on behalf of the Northern District of Texas, USA; has become a risky adventure for any lawyer. The judge doesn’t want to risk the “delusional creativity” of generative AI jeopardizing the smooth running of the trial, as recently occurred in a Manhattan court.

Although not formally cited by Judge Starr, last week’s case was “supplementary” to the legal investigation conducted by attorney Steven Schwartz through ChatGPT. To everyone’s surprise, The six argumentative situations presented were not real, but a “crazy” creation, that is, invented by the language model..

To prevent such a blunder on a national scale in his area of ​​expertise, the judge used his prerogatives as the owner of one of the country’s most important courts to create a “Mandatory certificate on productive artificial intelligence”. Must be filled by all lawyers.

What is Mandatory Certification for Generative Artificial Intelligence?

The model mandatory certificate designed by Judge Starr must be signed by all attorneys working in the court in question and filing with the case, and must certify that no part of it was written by artificially generative artificial intelligence (ChatGPT, Harvey.AI or Google Bard or similar). .

If some of the files have been prepared using such a device, petition authors also certifies that the “authenticity has been verified by a human using printed reports or conventional legal databases” of the content.

Regarding the reason why the model was created, the explanation given is, in the current format, these prolific AI platforms “make up everything from citations to references”. It also says in the explanatory note that while true lawyers swear to cast aside their prejudices, the same doesn’t happen with human-created programs without this concern.