AI and the law, part 2

« previous post | next post »

Here we go again, but this time on a grander and more dramatic scale:

Pras Michel of Fugees seeks new trial, contends former attorney used AI for closing argument

The hip-hop artist convicted on campaign finance and foreign influence charges seeks to set aside the jury’s guilty verdicts.

By Josh Bernstein, Politico (1016/23)

Notice the high stakes of this trial, since the defendant, among many other serious, wide-randing charges, is accused of acting as an unregistered foreign agent for China.

Fugees star Pras Michel, who was convicted in April on charges of conspiring to make straw campaign donations, witness tampering and acting as an unregistered foreign agent for China, appears to be breaking new legal ground by calling for a new trial by claiming his defense attorneys allegedly relied on artificial intelligence to compile their final argument for the jury.

In a withering motion filed Monday night with a federal judge in Washington, Michel’s new attorneys argued that his Los Angeles-based lawyer David Kenner relied on the fledgling technology at critical points in Michel’s trial, contributing to “prejudicial ineffective assistance of counsel.”

As soon as I saw David Kenner's name and photograph bruited in this case, I thought, "Isn't he one of the most prominent celebrity lawyers in LA?"

Indeed, he is.  See here and here.

Kenner “used an experimental artificial intelligence (AI) program to draft the closing argument, ignoring the best arguments and conflating the charged schemes, and he then publicly boasted that the AI program ‘turned hours or days of legal work into seconds,’” Michel’s new defense team from D.C.-based ArentFox Schiff wrote. “It is now apparent that Kenner and his co-counsel appear to have had an undisclosed financial stake in the AI program, and they experimented with it during Michel’s trial so they could issue a press release afterward promoting the program — a clear conflict of interest.”

“Michel’s prior publicist also informed members of the current defense team that Kenner proudly stated at the end of the trial words to the effect of ‘AI wrote our closing,’” former federal prosecutor Peter Zeidenberg wrote in a declaration accompanying the motion.

Zeidenberg pointed to a press release a firm called Eyelevel appears to have issued in May, which included a photo of Michel and boasted that the company’s technology “made history last week, becoming the first use of generative AI in a federal trial.” The release quotes Kenner calling the AI tool “an absolute game changer for complex litigation.”

The announcement does not mention that, after a three-week trial featuring a diverse set of witnesses ranging from Hollywood star Leonardo DiCaprio to former Attorney General Jeff Sessions, jurors took less than three days to convict Michel on all 10 felony charges he faced.

Federal prosecutors claimed Michel took staggering sums of money — about $88 million — from wealthy Malaysian businessman Jho Low in a bid to enhance his influence in the U.S. and later to derail the investigation and prosecution of Low on fraud charges stemming from the collapse of Malaysian sovereign wealth fund 1MBD. The hip-hop star who came to prominence in the 1990s was accused of arranging to donate money from foreign nationals to President Barack Obama’s reelection campaign in 2012 and making intensive efforts to influence President Donald Trump’s administration to drop or resolve the charges against Low.

The new lawyers also say Kenner was “deficient” by failing to seek separate trials for Michel on the straw donation charges related to the Obama campaign and the foreign-influence related charges from the Trump era. That decision may have been influenced by a “flat fee” Kenner agreed to with Michel, the new-trial motion says.

“Michel’s counsel was deficient throughout, likely more focused on promoting his AI program and saving himself from the contempt proceeding than zealously defending Michel,” Zeidenberg wrote.

Another article on the case provides additional details:

Star Hip Hop Lawyer Accused of Using AI in Big Case He Lost

He is also accused of violating a court order by sharing confidential information with the media.

Noah Kirsch, Daily Beast (10/17/23)

One wonders how Kenner ever climbed to such heights of the legal profession if he were so sloppy and lazy as he seems to have been in this case:

Star lawyer David Kenner—known for representing hip hop artists Snoop Dogg and Suge Knight—is accused of making several giant lapses during the conviction of his latest major client.

In a motion for a new trial filed on Monday evening, new attorneys for Fugees rapper Pras Michel blasted Kenner as unprepared and unscrupulous. Among their most startling claims: that Kenner used an “experimental” artificial intelligence program to help draft his closing argument in the trial. They said the program didn’t even perform its job well.

Later, the new defense team said, they became aware that Kenner had “a financial interest in the AI program.” He even offered a quote for a press release promoting the new technology, saying, “This is an absolute game changer for complex litigation…The system turned hours or days of legal work into seconds. This is a look into the future of how cases will be conducted.”

Michel was convicted in the spring on 10 counts, including for “orchestrating an unregistered, back-channel campaign” to try to persuade the Trump administration to drop an inquiry into a Malaysian businessman, Jho Low, who was accused of perpetrating a massive fraud scheme. Michel was also accused of working with Low on improper political contributions and lying to the Federal Election Commission about it.

His attorneys now assert that, because of Kenner’s allegedly poor representation, he “never had a chance.”

The implications of this case are vast.  Not only did Kenner admit that he used AI to prepare his case, he bragged about it, as though it were some positive, new development in the practice of law.

Selected readings

[Thanks to Kent McKeever]


  1. NSBK said,

    October 19, 2023 @ 3:17 pm

    I imagine that use of an LLM of some kind, trained on primarily legal documents and court proceedings, really could be beneficial to attorneys. In this particular case it seems like the bigger issue was that the attorney had a conflict of interest owing to his financial stake in the tool he used.

    But now a vision of a kind of legal distopia has been conjured in my head, where AI is also used by legislators to write new laws, causing an arms race that results in all legal dealings requiring the use of an AI, because humans can't keep up with the rising complexity of the law.

  2. Benjamin Ernest Orsatti said,

    October 19, 2023 @ 3:46 pm

    NSBK said, "I imagine that use of an LLM of some kind, trained on primarily legal documents and court proceedings, really could be beneficial to attorneys."


    No, it could not.

  3. Haamu said,

    October 20, 2023 @ 12:41 am

    @Benjamin Ernest Orsatti: Too late.

  4. Philip Taylor said,

    October 20, 2023 @ 4:46 am

    Disclaimer — I am not a lawyer, while I have every reason to believe that Benjamin Orsatti is. Nonetheless, when Benjamin writes (paraphrase) "use of an LLM of some kind, trained on primarily legal documents and court proceedings, really could not be beneficial to attorneys", I have some reservations. Let me start by saying that I regard it as beyond dispute that LLMs, at least their current instantiations, have zero awareness of the meaning (qua meaning) of anything that they churn out. However, imagine a generative AI LLM being prompted along the lines of the following : "Prepare a legal submission for the XXX court on behalf of Mr YYY (defendant) who is accused of ZZZ. Base your argument on the findings of the AAA court in the case of Mr BBB (defendant) who was similarly accused of ZZZ, and cite in support of your aergument only genuine cases that are recorded in <list of reliable legal databases>". If the output resulting from that prompt were then properly vetted by a qualified lawyer (or even para-legal), charged with ensuring that every element was factually correct, would that not satisfy Benjamin’s reservations while at the same time saving considerable time for the defendant’s legal team and thereby keeping his costs down ?

  5. Benjamin Ernest Orsatti said,

    October 20, 2023 @ 7:17 am


    Congratulations, you've invented Westlaw! (sorry).

    Westlaw and Lexis already sort of due this with varying degrees of success. You search the "Pa.Appellate.Cases.State" database for cases "tagged" with "real property tax appeal" and filter for the cases where the Plaintiff won.

    So, yeah, you're right, that sort of thing — when I'm plugging in the search string using Boolean terms and I'm controlling the database selection and all the filters, yeah, I'm not threatened by the Terminator robots there, I'll give you that one.

  6. Benjamin Ernest Orsatti said,

    October 20, 2023 @ 7:23 am

    Haamu said,

    "Too late".

    Maybe, but so far, Lexis & Westlaw's "AI" stuff — "brief analysis", "related content", etc., are about at the level that my 11 year-old could churn out.

  7. Gregory Kusnick said,

    October 20, 2023 @ 10:54 am

    "about at the level that my 11 year-old could churn out"

    And in fact if Kenner had paid an 11-year-old to write his closing argument, and used it unedited in court, and boasted about it afterward, he'd be just as guilty of legal malpractice as in the actual case. So I don't see that AI is really the issue here.

  8. Haamu said,

    October 20, 2023 @ 12:18 pm

    I can't speak to what Lexis & Westlaw are offering, but an 11-year-old who could write like any of the current prominent LLMs would be a scary thing indeed.

RSS feed for comments on this post