SHARE

In 2020, the average median income for a lawyer in the US was just under $127,000. All that money obviously has to come from somewhere—most often, the clients they represent. Unfortunately, many people neither have the budget for high-priced attorneys, nor the time to deal with the tedious red tape on their own. According to one startup, DoNotPay, often winnable appeals fall by the wayside, and they’re hoping to level the playing field a bit.

Since 2015, DoNotPay has offered increasingly nuanced and diverse legal advice via AI software trained on copious amounts of past court cases and law data. Last month, the startup’s latest toolkit update included the abilities to negotiate lower bills and cancel unwanted subscriptions while sparing consumers from lengthy customer service interactions. Taking things one step even further in 2023, DoNotPay now seeks to aid defendants in a real-life court setting.

According to New Scientist, the AI chatbot developers recently announced plans to supply an unnamed individual with a smartphone connected to their new program for their upcoming court date. The AI will then listen to the hearing’s proceedings as the defendant contests a speeding ticket, and supply them with every response to provide the judge via an earpiece.

[Related: OpenAI’s new chatbot offers solid conversations and fewer hot takes.]

If such a strategy sounds legally dubious, well—you’re probably right in most places. That said, DoNotPay’s CEO Joshua Browder, told New Scientist that they were able to identify a (currently unspecified) location of the country where listening via the earpiece is “technically within the rules,” albeit not in the “spirit of the rules.” In any case, the test case’s defendant doesn’t have much to worry about—regardless of outcome, DoNotPay agreed to pay their speeding fees if they happen to lose their appeal.

Generative AI programs have increasingly come to the forefront of artistic and ethical debates in recent years with the rise of projects like OpenAI’s impressive ChatGPT and Meta’s less-than-stellar BlenderBot 3 experiment. While it’s unsurprising to see the same kind of advancements find their way into legal settings, experts caution that we’re a long way from replacing all our legal professionals. “When your lawyer tells you ‘OK, let’s do A’, we trust them that they have the expertise and the knowledge to advise us,” Nikos Aletras, an AI designer at the University of Sheffield, UK, told New Scientist. “But [with AI], it’s very hard to trust predictions.”