{"id":9659,"date":"2025-11-06T21:23:29","date_gmt":"2025-11-06T21:23:29","guid":{"rendered":"https:\/\/journals.law.unc.edu\/ncjolt\/?p=9659"},"modified":"2025-11-06T22:12:38","modified_gmt":"2025-11-06T22:12:38","slug":"from-brainstorming-to-juror-ais-infiltration-into-the-courtroom","status":"publish","type":"post","link":"https:\/\/journals.law.unc.edu\/ncjolt\/blogs\/from-brainstorming-to-juror-ais-infiltration-into-the-courtroom\/","title":{"rendered":"From Brainstorming to Juror? AI\u2019s Infiltration into the Courtroom"},"content":{"rendered":"\n

4:24 PM, Nov. 6, 2025<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

Artificial Intelligence (\u201cAI\u201d) is changing the landscape of the legal profession. Recently, three generative AI Large Language Models (\u201cLLMs\u201d) participated as jurors in the Unprecedented Simulation<\/a>\u2014a simulated juvenile criminal trial\u2014begging the question about the validity of AI jurors. While asking \u201cif\u201d AI jurors are the future is loaded with ethical<\/a> and constitutional<\/a> implications, what about asking \u201chow?\u201d<\/p>\n\n\n\n

AI marches ever closer to the courtroom<\/a> with varying degrees of success<\/a>. Learning from the successes and mistakes of other AI tools may illuminate how to implement AI jurors.<\/p>\n\n\n\n

<\/p>\n\n\n\n

Existing Beneficial AI Tools<\/strong><\/p>\n\n\n\n

AI has become commonplace<\/a>. Generative AI streamlines<\/a> administrative processes across sectors. Industries such as healthcare<\/a>, business<\/a>, and law utilize AI tools by OpenAI\u2014a popular generative AI company. OpenAI has necessarily trained its LLM on a wide range of texts to benefit such a broad range of applications.<\/p>\n\n\n\n

With respect to law, practitioners utilize <\/a>OpenAI for legal research, document review, and drafting. Westlaw and Lexis both have AI research assistants<\/a> based on OpenAI<\/a>. Considering the available legal applications of OpenAI, it is unsurprising that ChatGPT<\/a> was one of the AI jurors<\/a>. In fact, the Unprecedented Simulation praised ChatGPT\u2019s suitability as a juror because of its broad applicability.<\/p>\n\n\n\n

<\/p>\n\n\n\n

AI Cautionary Tales<\/strong><\/p>\n\n\n\n

Generative AI requires caution. Westlaw and Lexis\u2019 generative AI tools<\/a> can assist with drafting, but these tools are not perfect<\/a>. Hallucinations<\/a> are a costly concern in the legal profession. Trial outcomes often turn on precedent and binding case law. Accordingly, established procedures incentivize<\/a> all litigants to adhere to courtroom standards.<\/p>\n\n\n\n

Court documents created by generative AI can be checked for validity, and generative AI can be given different prompts to course correct any outputs. However, jury rooms are black boxes<\/a>. The system is not designed to accommodate checkpoints or oversight. <\/p>\n\n\n\n

Instead of plug-and-playing jurors with prefab generative AI, agentic AI may be a better fit.<\/p><\/blockquote>\n\n\n\n

Generative vs. Agentic AI<\/strong><\/p>\n\n\n\n

Instead of plug-and-playing jurors with prefab generative AI, agentic AI may be a better fit. Generative AI is \u201creactive to the users’ input<\/a>.\u201d It primarily responds to prompts by creating \u201cnew content<\/a>\u201d based on its training sets.<\/p>\n\n\n\n

Alternatively, agentic AI is more \u201cproactive<\/a>.\u201d Agentic AI is intended to \u201cmake decisions based on context<\/a>\u201d and accomplish \u201cmultistep processes<\/a>.\u201d That\u2019s not to say that agentic AI is a better product than generative AI; agentic AI still needs oversight<\/a> and is vulnerable to error<\/a>.<\/p>\n\n\n\n

<\/p>\n\n\n\n

The Unprecedented Simulation<\/strong><\/p>\n\n\n\n

When the AI jurors deliberated, each AI gave a comprehensive retelling of law and facts. The amount of specific recollection was impressive<\/a>. But as the rounds of deliberation progressed, it seemed like all three AI got stuck in a predictive loop<\/a> exploring only one line of inquiry\u2014mere presence versus accomplice liability\u2014instead of returning to other lines of inquiry. For example, the AI jurors seemingly ignored whether the defendant knew the other people involved in the crime, failing to consider making an inference based on the circumstances.<\/p>\n\n\n\n

Because juries function as decision makers<\/a> instead of content creators, agentic AI presents an interesting option to star as jurors. Juries are essentially entities who make decisions based on the context of the facts. Factor tests or element tests in law require multistep processes or analyses to distill into a single answer. Furthermore, juries are not creating new content, only deciding if the evidence proves the prosecutor\u2019s case beyond a reasonable doubt (in the criminal context).<\/p>\n\n\n\n

Ultimately, technology is not quite ready<\/a> to release AI jury trials as a viable alternative to human jury trials. Instead of comparing AI jurors to human jurors, maybe AI could help judicial economy stretch farther by completely avoiding trials (e.g. by resolving prosecutions or disputes via plea negotiations or settlements<\/a>). Perhaps an agentic AI designed to meet that standard could bring about the Next Simulation.<\/p>\n\n\n\n

<\/p>\n\n\n\n

Eli Lynn<\/strong><\/p>\n\n\n\n

Eli graduated from Campbell University with a B.S. in Psychology. In law school, Eli is the President of Parents as Law Students, Treasurer of OutLaw, Treasurer of the UNC chapter of the National Lawyers Guild, Certified Law Student with the Criminalized Survivor, Detention, and Justice Clinic, and competing member of Holderness Moot Court\u2019s Appellate Advocacy team. Depending on the weather, Eli can be found poolside or inside. Puzzles and the Carolina Hurricanes carry them through the winter.<\/p>\n","protected":false},"excerpt":{"rendered":"

Instead of plug-and-playing jurors with prefab generative AI, agentic AI may be a better fit<\/p>\n","protected":false},"author":4,"featured_media":9661,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":[],"categories":[51],"tags":[416,297,666,333,505,669,504,470,667,188],"_links":{"self":[{"href":"https:\/\/journals.law.unc.edu\/ncjolt\/wp-json\/wp\/v2\/posts\/9659"}],"collection":[{"href":"https:\/\/journals.law.unc.edu\/ncjolt\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/journals.law.unc.edu\/ncjolt\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/journals.law.unc.edu\/ncjolt\/wp-json\/wp\/v2\/users\/4"}],"replies":[{"embeddable":true,"href":"https:\/\/journals.law.unc.edu\/ncjolt\/wp-json\/wp\/v2\/comments?post=9659"}],"version-history":[{"count":4,"href":"https:\/\/journals.law.unc.edu\/ncjolt\/wp-json\/wp\/v2\/posts\/9659\/revisions"}],"predecessor-version":[{"id":9670,"href":"https:\/\/journals.law.unc.edu\/ncjolt\/wp-json\/wp\/v2\/posts\/9659\/revisions\/9670"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/journals.law.unc.edu\/ncjolt\/wp-json\/wp\/v2\/media\/9661"}],"wp:attachment":[{"href":"https:\/\/journals.law.unc.edu\/ncjolt\/wp-json\/wp\/v2\/media?parent=9659"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/journals.law.unc.edu\/ncjolt\/wp-json\/wp\/v2\/categories?post=9659"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/journals.law.unc.edu\/ncjolt\/wp-json\/wp\/v2\/tags?post=9659"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}