Judicial guidanceRevised October 14, 2026

Responsible AI Use in Litigation

Judge Maritza Dominguez Braswell’s practical guidance on verification, discovery, confidentiality, and correcting AI errors.

Guidance · Existing rules still govern

Why I’m including this

A practical guide to responsible use

I’m including this guidance because it gives lawyers a clear, practical approach to using AI in litigation. Judge Braswell explains how the technology works, keeps professional judgment and verification central, and offers a constructive process for correcting errors. Her emphasis on education, accountability, and prompt correction makes this a useful reference for everyday practice.

Understanding the technology · p. 1

What the guidance means by AI

As used here, “AI” refers to generative AI (GenAI): tools that produce text, images, audio, or video in response to a user’s instructions, often given in everyday plain language. ChatGPT, Claude, and Gemini are examples of GenAI tools developed by frontier AI labs. It is important to know that AI is often built into document review software, word processors, phones, and more. So even if you never open a tool like ChatGPT or Claude, you may still be using AI.

Judge Braswell’s definition · p. 1

Probabilistic vs. deterministic systems · pp. 1–2

Predicting a response differs from calculating or retrieving

Judge Braswell explains the distinction through familiar tools:

How calculators, databases, and generative AI produce results
System	Example	How it produces a result
Deterministic	Calculator	Applies fixed rules to an input. The same calculation produces the same result.
Retrieval	Traditional database	Returns an existing document.
Probabilistic	Generative AI	Uses learned patterns to predict a likely response, piece by piece. The same question can produce different answers.

A fluent answer is not proof of accuracy. Legal AI tools that link to source databases still require verification.

Read the distinction · pp. 1–2

The judge’s framing · p. 2

Why these practices reach beyond your own AI use

The practices below are not legal advice, nor are they directives. Still, you might consider applying some or all of them in light of your existing litigation and/or professional obligations. Moreover, because AI is increasingly integrated into many everyday tools and can enter a case in several different ways, some of these principles and practices may be relevant to your work even if you do not use AI tools yourself.

Judge Maritza Dominguez Braswell · p. 2

Practice 2 · p. 3

Keep your judgment in control

Form your own views before using AI. Verify the final work. Source

Practice 3 · p. 3

Verify outside the tool

Check authorities against reliable sources. Asking AI to confirm itself is not verification. Source

Practices 4–6 · p. 3

Match review to risk

Apply greater scrutiny to court submissions. Set guardrails and keep filings concise. Source

Practice 7 · p. 4

Check confidentiality terms

Understand retention and training settings. Confer early about protective orders. Source

Practices 8–9 · p. 4

Plan discovery around AI

Discuss AI use with clients and experts. Address preservation and discovery protocols early. Source

Practices 10–11 · pp. 4–5

Be candid about material uses

Raise supported concerns promptly. Identify AI use affecting evaluation of evidence; explain legitimate disclosure concerns. Source

Correcting false citations · p. 5

Correct, confer, and inform the court

For your own error, promptly correct it and notify the court and opposing counsel.

For an opponent’s error, the judge suggests conferral, joint notice, and escalation if necessary.

Read the correction framework