Judicial AI Task Force to release key guidance for California Courts on Feb. 21

This Friday, Feb. 21, California's Judicial Council will receive a 30-minute presentation on the "Model Policy for Use of Generative Artificial Intelligence," developed by the Judicial Branch Artificial Intelligence (AI) Task Force. While no action or report is expected at the meeting, this model policy marks a significant step in California's judiciary addressing the safe and responsible use of generative AI in court administration. Courts reportedly will have the flexibility to adopt or modify the policy as needed, with an initial focus on court administration.

The AI Task Force's work extends beyond governing court administration or operations. It also plans to develop a rule of court for adopting AI use policies and a standard of judicial administration governing AI use by judicial officers. Notably, external uses of generative AI, such as customer service chatbots, are not currently addressed. Internal uses of generative AI are referenced. Omitting the external use of AI raises questions about any potential bans on such applications, which are widely adopted in private industries to enhance efficiency and accessibility in connection with customer service.

California's courts have been awaiting comprehensive guidance on AI's role in legal proceedings, particularly regarding its use by judges, court personnel, lawyers, and litigants. It's been 27 months since GPT 3.5 appeared and 14 months since Chief Justice of the United States John G. Roberts, Jr. focused on AI in his annual report on the judiciary in 2023. California Chief Justice Patricia Guerrero emphasized AI as a branch priority in her March 2024 State of the Judiciary Address. The AI Task Force, composed of appellate justices, trial court judges, an attorney, and a court administrator, is now poised to deliver initial guidance. However, its work will remain ongoing, reflecting the fast-evolving nature of AI technology upending the legal profession and law practice.

California's judiciary appears unlikely to ban AI outright, as no state judiciary has attempted such a prohibition. Instead, other states have emphasized the "safe and responsible" use of AI, placing accountability on users. States like New York, Illinois, Delaware, and Kentucky have issued interim guidance addressing key concerns, such as judicial ethics, confidentiality, and impartiality. California's AI Task Force will likely follow suit addressing challenges posed by generative AI. Since the formation of the AI Task Force, AI technology has rapidly advanced to include Agentic AI. It is doubtful the AI Task Force has been able to keep up with current AI developments so more work will likely follow potentially in perpetuity.

At a minimum, the AI Task Force is expected to highlight problematic areas for courts to consider when implementing AI. The National Center for State Courts (NCSC) has been a resource for courts nationwide, offering state-of-the-art AI information since 2023. California's approach will likely echo and possibly expand on these efforts.

Despite the lack of formal guidance to date, some California courts have already adopted AI tools. Orange County Superior Court uses internal chatbots, EVA and EMI, to assist court clerical staff with training and HR inquiries. These chatbots, designed in-house, operate securely on court-specific cloud servers, mitigating security and accuracy concerns as they were trained on internal materials. Orange County is sharing the architecture for these tools with other California courts. Although California has a unified statewide branch, each of California's 58 counties operates its own Superior Court with local control, meaning they each purchase their own case management systems and modernize technologically as statewide funding allows.

Los Angeles County Superior Court, in collaboration with Stanford Law School, is leveraging AI to address access-to-justice challenges. Focusing on limited civil cases, such as unlawful detainer and debt collection, Stanford Law School's initiative with the nation's largest trial court aims to refine court processes and potentially reduce unmeritorious default judgments against self-represented litigants according to Professor David Engstrom. By identifying common procedural gaps, the project seeks to enhance court efficiency in identifying incomplete default judgment requests and ensure self-represented litigants receive fair treatment. Professor Engstrom sees Court AI as a future remedy to the civil justice crisis facing America.

The introduction of AI into the judiciary raises critical ethical and procedural questions. The greatest concern identified nationally is the improper delegation of judicial decision-making. There is apprehension judges will allow AI to render final decisions, even though the use of AI to supplant human judgment would constitute a dereliction of judicial duty under current ethical norms.

Judicial ethics codes, such as the California Code of Judicial Ethics and the Model Code of Judicial Conduct, already impose constraints that are highly relevant to AI use. Key issues include:

Ex Parte Communications: Could a judge's use of AI trained on extraneous information violate prohibitions against ex parte communications?

Confidentiality: Should judges be restricted to AI housed on private servers to avoid breaches of sensitive information?

Impartiality: Can AI systems trained on biased data compromise judicial neutrality?

Transparency: If judges rely on AI, will parties have the right to know how decisions were influenced by the technology?

These concerns underscore the need for clear guidance to ensure that AI use aligns with the judiciary's ethical obligations.

Transparency is a cornerstone of the justice system, yet proprietary AI systems operate as "black boxes," making it difficult to assess their decision-making processes. Additionally, AI's inherent biases - stemming from training data that reflect societal prejudices - pose significant risks to fairness within the judicial sphere. Mitigation strategies will be essential to ensure AI's compatibility with the judiciary's principles of impartiality and equal justice.

Another pressing issue is the risk of "hallucinations" by AI, where systems generate false or misleading information. While trial courts often render decisions that are later overturned, reliance on AI-generated errors could undermine public trust in the judiciary. Courts must also address the challenge of fabricated evidence, such as deepfakes. The sophistication required to create convincing falsified evidence has decreased dramatically, making it imperative for courts to establish protocols to detect and exclude such materials immediately. It is not clear that the AI Task Force will imminently address practical concerns such as deepfakes in evidence daily received by trial judges in family, civil, criminal, probate or other such cases.

The AI Task Force is unlikely to recommend sweeping restrictions on AI. Instead, its guidance will likely focus on balancing innovation with caution, emphasizing accountability, transparency, and ethical use. Court administrators and judges will need to remain firmly in control of decision-making, using AI as a tool rather than a substitute for human judgment. Whether judicial officers will need to disclose AI use on the record remains an open question.

California's approach to AI in the judiciary will have far-reaching implications, not only for the state but also as a model for other jurisdictions navigating similar challenges. By establishing thoughtful, forward-looking policies, the Task Force can help the judiciary harness AI's potential to improve access to justice, enhance efficiency, and uphold the rule of law. If the AI Task Force restricts the implementation of modern technology local trial courts wish to implement, it risks re-opening old battle lines about the independence of trial courts dating back to unification permitted by California's voters in 1998 through a constitutional amendment. Hopefully, the AI Task Force provides sufficient useful guidance, while allowing for ongoing local innovation unhindered by statewide mandates.