AI foundations in the courts

Larger organizations are often able to afford advanced AI tools, either built or bought, while smaller organizations may struggle to keep up. Courts with limited budgets can look for more cost-effective solutions and consider collaborative efforts or grants to access AI technologies. It's also important to train team members on the effective and efficient use of AI to avoid unexpected costs.  This webinar AI in Action:  Current applications in state courts examines how small and large court systems are using AI.

Access to justice is an ongoing concern, and legal aid organizations are leaders in the use of AI in the legal industry.  Check out the case studies in the webinar Tech for All: Applications of AI to increase access to justice. 

Please refer to our December 2024 webinar, AI in Action – Current Applications in State Courts, for a discussion on this issue.

AI and Robotic Process Automation (RPA), separately or in combination, can significantly streamline paperwork. It automates the extraction of both structured and unstructured data from forms, then efficiently populates databases or other digital systems with that information, reducing manual effort and potential errors. AI is good with natural language processing and with image or document recognition. RPA is good at data entry automation. 

Generative AI tools are not intended to replace judges, lawyers, or human decision-making in the court system (view this video to see why). Rather, they are designed as assistants to enhance efficiency and provide support. To learn about these potential use of AI and its limitations, view Fundamental of AI in the U.S. Court system and A Legal Practitioner's Guide to AI and Hallucinations

Consider these points:

• Generative AI is not a decision-maker but is instead designed for research and analysis, similar to databases and other online legal research tools. It can quickly process vast amounts of information, suggest relevant precedents, or draft initial documents, but all final decisions should rest with humans.
• AI augments human expertise but doesn't replace it. AI can help legal professionals work more efficiently by automating routine tasks, freeing them to focus on analysis and decision-making that require human judgment.
• Any output from AI tools should always be reviewed, verified, and approved by qualified legal professionals before being used in court proceedings or client interactions. This ensures that information is accurate, relevant, and presented within the correct context.
• As AI technology evolves, it is increasingly able to process and analyze complex information. However, it is crucial for courts to regularly assess and update these tools to ensure they are accurate and reliable.
• AI use in legal settings is the subject of ongoing ethical discussions and regulations to ensure fairness, transparency, and accountability.  To learn more about ethical uses of AI in legal and court settings, view Navigating AI in court settings, Ethics, Legal Frameworks, and Practical Tools.
• AI does not "reason" in the same way humans do, but when used responsibly along with human expertise, it can be a valuable tool to enhance the efficiency and effectiveness of any legal professional. The key is to view AI as a supportive technology rather than a replacement for human judgment.

Your court should establish clear guidelines, develop evaluation processes, and ensure that all AI tools have human oversight. This includes comprehensive bias testing and ensuring your AI output aligns with ethical standards. In addition, creating and rigorously enforcing rules that govern acceptable AI use by court professionals, collaborating with AI developers, and continuously monitoring your AI can help ensure ethical use. Use this AI readiness guide developed by court professionals to get started.

Responsible use of AI in your court requires strict adherence to privacy laws, ethical guidelines, and sustained monitoring. You should implement robust testing procedures and ensure that AI inputs and prompts are secure and private and that AI outputs are accurate, unbiased, and supervised by humans. Collaboration with experts such as ours can also enhance responsible AI use. Our Principles and Practices for AI in State Courts provides a self-assessment for AI use. 

Transparency is critical in AI development. Your court should demand clear information from AI providers about how data is used and how models are trained. Open communication and transparency are foundations that help build trust and ensure that ethical standards are met. Courts must also be transparent with court users about any use of AI that may affect them. This AI readiness guide for courts provides details of how to administer effective AI governance.

Oversee an AI system the same way you would evaluate someone working in a job. Ask specific questions to detect potential biases and continue regular assessments of AI output to check for accuracy and completeness. You can also develop a set of specific test cases, set clear performance guidelines, and be ready to stop using AI tools that demonstrate bias. You can constrain AI to work only with the documents you give it and also choose a system that uses Retrieval-Augmented Generation (RAG) so your AI consults a trusted knowledge base rather than scanning the entire Internet. To better understand how courts are using these advanced techniques, view AI in Action:  Current applications in state courts.

Large language models are typically trained on vast amounts of data, which can include biased or inaccurate information. That's why it's important to use AI as a tool, be aware of its limitations and potential for bias, and verify any AI-generated information against other reliable sources. To learn about these potential benefits and limitations, view Fundamentals of AI in the U.S. Court system.

Choosing service providers that honor data privacy requirements and minimize data leakage is a good way to start. You can also use a local VPN or Large Language Model, dedicated endpoints, and user authentication tools. Local versions of AI that don't send data to external servers and options that keep data from being used to retrain models can also help preserve privacy.  Learn more about the ethics of the use of AI in courts and current legal frameworks and practice tools.

Clarity, specificity, and context. You want Generative AI to understand the task. If you are asking about a legal scenario, include the jurisdiction, area of law, and pertinent case details or legal principles.  View this webinar to learn more best practices about prompt engineering.

Here are some strategies for better prompts:

• Try using a back-and-forth approach, also known as iterative refinement. To do this, start with a basic prompt and then ask follow-up questions or provide instructions to delve deeper or correct misunderstandings.
• Be aware that your AI tool has a "context window," and once you reach the limit of the context window, the AI will "forget" what it has been told previously in the chat. If you encounter this issue, you can remedy it by including the relevant details from earlier in the chat in your next prompt.
• You can also ask the AI how to craft an effective prompt for the task. Known as "meta-prompting," this technique can be powerful because it leverages the AI's knowledge of concepts and terminology.
• Take time to experiment with different phrasings and approaches. What works well for one task may not be ideal for another, so it's worth it to develop prompt templates to help handle your common workflows.

Yes, and that's why all AI output should be verified. If a GenAI tool does not have access to the relevant data or if you have not given a specific, clear prompt with the proper context, it may guess what you want. AI answers can still sound authoritative even if they're incorrect. Examples include citing case law that does not exist, and this kind of false information is referred to as "a hallucination."

Read  A Legal Practitioner's Guide to AI and Hallucinations for additional guidance.

For document summarization, the time savings comes from a fundamental difference: verification is faster than creation from scratch.

When verifying AI-generated summaries, you're reviewing a structured synthesis with specific claims to check – not reading dozens of documents without knowing what's important. The AI handles the heavy lifting of processing large volumes quickly, identifying key points, and creating an initial structure.

The real efficiency gain is shifting from "read everything → synthesize" to "review AI synthesis → verify key points." Even with thorough verification, checking a structured output against sources is typically much faster than creating that output yourself. AI tools with citation features make this process even more efficient by letting you spot-check claims and trace assertions back to source material without rereading everything.

The value proposition is strongest with large document volumes. For very short documents, or when deep domain expertise is needed, there may be no advantage to using AI.

Courts typically secure funding through a combination of approaches. This includes integrating AI development into their existing technology budgets and actively pursuing grant opportunities from sources like bar associations, local foundations, or specialized legal technology funds.

Yes, security is paramount. Your court or organization must have a robust data governance policy specifically addressing the security of sensitive information. When using AI, never input sensitive client data into public-facing AI systems. The best practice is to implement a Large Language Model (LLM) within your own secure, private environment for staff use, ensuring all data remains under your direct control and is protected.

Ensuring accessibility for all users, including those with disabilities, is a critical commitment. Regardless of whether AI is involved, courts are obligated to ensure their websites and digital tools are fully ADA compliant. Courts should adhere to established guidelines, such as those found on section508.gov, to provide accessible design and functionality for all users.

AI-assisted translation and transcription tools show promise for improving language access in courts, but they cannot replace qualified human interpreters — particularly in courtroom settings. Machine translation has not developed to the point where it can be relied on to convey complex information, legal concepts, or information that could impact people's ability to understand and exercise their legal rights. Courts should never use machine translation for court events, to convey legal or procedural information, or to carry out complex, extended interactions. However, courts may be able to use machine translation for limited out-of-court interactions or to create drafts of documents to be reviewed by professional translators.

When AI translation tools are used, a human-in-the-loop approach is essential: the U.S. Department of Justice guidance discourages the use of machine translation alone even with the use of a disclaimer, and recommends that any machine-translated information be reviewed by a professional translator.

For deeper guidance, see NCSC's resource Machine Translation: Considerations and Cautions for Courts, and the TRI/NCSC AI Policy Consortium webinar AI-Assisted Translation in the Courts: Exploring Opportunities & Navigating Risks.