A Warning for Clients, Lawyers, and Expert Witnesses
Artificial intelligence has rapidly become part of modern litigation. Lawyers use AI to summarize records, draft outlines, identify legal issues, and assist with research. Expert witnesses use AI to organize data, locate documents, and sometimes help analyze large volumes of information. Clients increasingly use AI to understand their legal claims, prepare timelines, or communicate with counsel.
But a new legal question is emerging:
Can the prompts entered into an AI system be compelled in discovery?
A recent federal court decision suggests the answer may be yes.
The First Major AI Prompt Discovery Battle
In May 2026, a federal magistrate judge in Connecticut (See Conservation Law Foundation v. Shell Oil Co.. Civil No. 3:21-cv-00933 (VDO) (D. Conn) ordered a party’s expert to produce the AI prompts used by an expert witness in preparing her expert report. The expert had used generative AI to help identify and sort potentially relevant documents from a large production of evidence.
The court concluded that the expert's AI prompts were potentially discoverable because they formed part of the expert's methodology. In other words, if an expert relies on AI to reach an opinion, the opposing party may be entitled to examine how that AI was used and what instructions were given to it.
The decision is currently under review, but it represents a significant development in the law of discovery and expert testimony.
Why AI Prompts Matter
Many people think of AI prompts as nothing more than questions asked to a computer.
In reality, prompts often reveal:
- What information the user considered important.
- What assumptions were made.
- What facts were emphasized or ignored.
- Whether the user was seeking objective analysis or a predetermined answer.
- The sequence of reasoning used to reach a conclusion.
For example, consider these two prompts:
Prompt A:
"Analyze all medical records and identify evidence supporting traumatic brain injury."
Prompt B:
"Analyze all medical records and identify evidence supporting and refuting traumatic brain injury."
The second prompt is objectively more balanced. The first may suggest confirmation bias. If an expert relies heavily on AI-generated analysis, an opposing attorney may argue that the prompt itself reveals a flawed methodology.
Why This Matters for Expert Witnesses
Experts have always been required to disclose the basis for their opinions.
Traditionally, opposing counsel could inquire about:
- Testing performed.
- Scientific methods used.
- Literature reviewed.
- Calculations performed.
- Assumptions relied upon.
Now courts are beginning to view AI prompts as simply another component of an expert's methodology.
An expert who uses AI may eventually be asked:
- What AI platform was used?
- What prompts were entered?
- Were prompts revised?
- What outputs were generated?
- Were any outputs discarded?
- Did the expert independently verify the AI's conclusions?
These questions may become common topics during expert depositions.
Why This Matters for Lawyers
Lawyers face a different set of concerns.
Suppose a lawyer uploads facts into an AI system and asks:
"Find weaknesses in my client's case."
That prompt may reveal legal strategy and attorney mental impressions.
Historically, such information would often be protected by the attorney work-product doctrine.
Whether AI prompts remain protected may depend on:
- The AI platform used.
- Whether confidentiality protections exist.
- How the information was stored.
- The purpose for which the AI was used.
- The jurisdiction involved.
Courts have not yet reached a consensus.
Why This Matters for Clients
Clients should understand that communications with AI are not necessarily private merely because they occur on a computer.
Consider a personal injury claimant who asks AI:
- "How much is my case worth?"
- "How can I make my injuries sound more serious?"
- "What facts are important to mention?"
Those prompts may create records that could later become relevant if litigation arises.
Similarly, businesses increasingly use AI to evaluate claims, investigate incidents, or analyze liability exposure. Those prompts and outputs may someday become discoverable evidence.
Imagine a client charged with a crime. They put into the AI prompt “How do I get away with murder” or “How do I successfully plead insanity”? If the prosecutor discovers these prompts, they will want to use them against the defendant.
The safest assumption is that anything entered into an AI system could potentially become part of future litigation.
The Risk of AI Hallucinations
Another concern is the well-publicized phenomenon of AI hallucinations.
AI systems sometimes generate information that sounds persuasive but is entirely incorrect. Courts across the country have sanctioned attorneys for filing briefs containing fabricated cases generated by AI.
As a result, lawyers and experts should never rely upon AI-generated information without independent verification. AI should be treated as a tool, not as an authority.
Practical Tips for Lawyers
Law firms should consider implementing written AI policies that address:
- Confidentiality and privilege concerns.
- Client consent where appropriate.
- Approved AI platforms.
- Record retention practices.
- Verification requirements.
- Documentation of AI-assisted work.
Attorneys should assume that future courts may scrutinize AI usage in the same manner they examine traditional expert methodologies and discovery practices.
Practical Tips for Expert Witnesses
Experts using AI should consider:
- Maintaining records of prompts and outputs.
- Documenting independent verification steps.
- Preserving information sufficient to explain methodology.
- Avoiding prompts that suggest a predetermined outcome.
- Being prepared to defend AI-assisted processes during deposition.
Transparency and reproducibility are likely to become increasingly important.
Looking Ahead
The legal profession is entering a new era. Artificial intelligence offers tremendous opportunities to improve efficiency and reduce costs. Yet every technological advance brings new legal questions.
The emerging dispute over AI prompts illustrates a larger reality: when AI becomes part of the decision-making process, courts may eventually demand to see behind the curtain.
For personal injury clients, businesses, lawyers, and expert witnesses alike, the safest course is to assume that AI usage may someday be examined by a judge, jury, or opposing counsel.
The lesson is simple:
Use AI thoughtfully. Verify everything. And never assume your prompts will remain hidden forever.
FINAL THOUGHTS:
1: At a minimum, use the paid version of an AI. The free version is more likely to be free game on discovery. The paid version is not a guarantee, but better than the free version.
If you are a client, do not use AI unless you are comfortable with a court possibly seeing your search terms.
If you are an expert witness, if you consult AI specifically for the case you are working on, there is a good argument that your prompts and other are discoverable.
If you are a lawyer, I think the work product privilege protects your prompts, but we will see what court rulings ultimately say.
Just as zoom depositions after Covid issued in new rules, issues and procedures, so does evolving technology. Smart lawyers adopt and prosper. Those not using AI risk being left behind.
Adam Sorrells
Law Office of Adam Sorrells
Serving injury victims throughout California.