Is the U.S. headed towards an AI-driven "smart court," as the Center for Strategic and International Studies calls China's frequent use of automated, digitized court proceedings? Not quite, experts say. However, these predictions aren't entirely off the mark.
"AI is really reaching all aspects of the law," said Wayne Cohen, managing partner at Cohen & Cohen and a law professor at the George Washington University School of Law.
While the current use of AI in the U.S. legal industry operates intensely behind the scenes, it's inching further into the front lines of the courtroom.
Cohen said AI plays a role in most of the research, writing and jury exhibit creation that goes into trial preparation, as well as office administration, trial summaries and translations.
It also helps kick the can down the road when processing lawsuits. "The movement of the cases from when a party files a lawsuit until the case is resolved is going to get much shorter," Cohen said.
From the bench, judges can generate searchable PDF transcriptions from audio recordings and make informed judgments that day. And with AI's ability to flag contradictions, it can bolster or hinder the credibility of the prosecution or defense. When judges make rulings, "they can do it with a lot of accuracy, and it's supported by the evidence that they heard in their courtroom," said Jackie Schafer, a former assistant attorney general for the state of Washington.
Schafer founded Clearbrief in 2020, which runs on AI that's designed to scan documents and identify citations, in addition to creating hyperlinked chronological timelines of all of the dates mentioned in documents for swift reference.
Jason Boehmig, CEO and co-founder of digital contract company Ironclad and who has experience as a corporate attorney, said AI can review a company's legal contracts, learning its preferred language and drafting and negotiating contracts in the organization's historic legal voice.Â
Business contracts are at the forefront of legal innovation, Boehmig said. "It's an area where we can afford to experiment," he said. On the spectrum of the legal system, the businesses on either end of the contract arguably have less to lose than, say, an individual whose basic freedoms are at stake.Â
In all of these applications, experts say the ideal situation is for humans to review AI's work. The notion of keeping the human in the loop is far from unique to the legal industry, but the significant ramifications coming out of the justice system make human oversight all the more critical.
Can AI reduce bias and costs in the legal system?
In weighing what can be gained versus lost in the use of AI in the legal system, the improvement of access to justice and the simultaneous upholding of due process are at the forefront.
Statistics of bias have proven the U.S. justice arena to be imperfect, and disparities can be seen across ethnicities as well as across wealth classes. "Most people that need legal services don't have access to them," said Boehmig.
"AI is leveling the playing field because it's allowing the lawyer all kinds of support for a fraction of the cost of human capital," Cohen said.
The concept that the world of legal representation will pass along their savings to those they represent may be idealistic, but some experts take a hopeful view.
More affordable representation could become a greater threat in the courtroom. Cohen said, "When an individual sues a corporation, normally the corporation has much better access to better lawyers, more resources, better support. Now, it sort of tips the scales."
Schafer also notes that pro se litigants, or those representing themselves without the presence of an attorney, are able to use AI technology to appear more professional to judges and juries. "They can come more prepared, even if they don't have legal training," Schafer said.
The perils of inserting AI into the legal process
The expectation of due process makes accuracy with AI a pivotal question, hence the need for an "AI sandwich," as Cohen puts it, or humans on either side of AI outputs.
For example, Mata v. Avianca was a New York case decided this past June where lawyers submitted fake quotes and citations created by ChatGPT. The attorneys were fined for going rogue, and the legal industry got a clear message that such a manipulation of AI is unacceptable. "Ultimately, it's the lawyer's license that is on the line," Cohen said.
For U.S. laws, which are complexly layered with local, state and federal expectations, there's only so much AI will be able to accomplish in the near term. For example, Schafer envisions small-time traffic tickets that have digital evidence can potentially be automated, but anything that requires critical human judgment cannot.
Meanwhile, there is the concern that AI will cause bias to increase. While AI has the potential to increase access to justice for marginalized groups, it also has the capacity to introduce new bias or double-down on existing bias. For example, AI-generated defense or prosecutorial summaries carry risk if the person reading them (such as a jury member) does not understand what follow-up questions to ask or where gaps in the information may be. "It's a bad idea to include a summary without the right context around it," Boehmig said.
When it comes to AI, the legal industry "is late on these innovation cycles in part because it's rational and efficient for the law to be late on matters of justice," Boehmig said. "When someone's liberty is at stake, I think we should probably have a human there for a long time."
