Between two of the most distrusted classes of modern society—lawyers and robots—who do we have less faith in?

The fear that artificial intelligence could consume blue-collar jobs is as old as the technology itself. Now, it seems that threat is expanding to the white-collar realm, as an AI-based legal advisor is posed to serve as a lawyer in a real court case for the first time. The algorithm, created by the legal services chatbot company DoNotPay, will be employed through a smartphone in the defendant’s pocket, advising them what to say via an AirPod (allowed per an accessibility loophole).

While the circumstances of the case have been largely under wraps, it’s been disclosed that the chatbot lawyer will make its debut in traffic court, where the defendant will repeat its outputs verbatim—without the judge’s knowledge. DoNotPay has agreed to incur any costs faced by the defendant, in the case that its AI gets stage fright or proves otherwise unsuccessful in stepping up to the challenge.

This sort of technology isn’t new to the legal field per se; it’s been used to aid customers who are pressing claims with governments and corporations, and has even assisted legal teams in the discovery process of cases. But this will be its first attempt as a sole legal advisor in court. In a number of ways, AI has proven itself to be largely superior to human lawyers. Tested against a group of 20 lawyers from reputable firms like Goldman Sachs and Cisco, a study by LawGeex found that human participants took an average of 92 minutes to review nondisclosure agreements for risk with a mean accuracy rate of 85 percent, while AI received an accuracy score of 94 percent in just 26 seconds.

The nuances of arguing a court case are profound, in that the twitch of an eye can change a lawyer’s tactics, and how the emotional delivery of a closing argument can sway the jury—or, at least, so say many series from Hollywood. DoNotPay has already begun troubleshooting its tech, which was recently updated to allow a two-second window for its clients to reject suggestions—it previously had a tendency to lie in the effort to solidify a win. Its abilities are already incredible, especially in quickly parsing through time-consuming documents that eat up the very expensive time of legal teams. DoNotPay is hoping this trial case will be a step toward matching the abilities of an actual attorney, at a substantially lower price. If successful, it will pose a divisive question to the public: Between two of the most distrusted classes of modern society—lawyers and artificial intelligence—who do we have less faith in?