Litigation and expert support

Technical expertise for matters involving AI

When an AI system is part of the facts, counsel needs to know what it said, why it may have said it, and what the evidence can support. Modulus provides legal teams with independent technical expertise, documented evidence, and research, and helps them understand how AI systems work.

Expert witness

Independent technical experts

Disputes involving AI often turn on technical questions that are new to courts, arbitrators, and the parties themselves. Counsel needs an expert who can examine the facts about an AI system, explain them clearly, and stand behind an analysis that will be tested.

In litigation and other proceedings, Modulus experts provide independent technical reports and testimony on how AI systems work and what the evidence in a matter shows. Counsel reviews each expert’s qualifications and any potential conflicts before retention.

  • Early consultation as counsel assesses the technical questions in a matter
  • Independent technical reports on the AI systems and outputs at issue
  • Review of the technical evidence, methods, and reports offered by other parties
  • Testimony at deposition, hearing, or trial when a qualified expert is retained to testify

An expert’s opinion is the expert’s own. It is formed from the evidence and the methods applied to it, it is not advocacy for any party, and no opinion is promised before the analysis is complete.

Understanding AI

How AI systems represent people, companies, and brands

We brief legal teams in plain language on how these systems arrive at what they say about people, companies, products, services, and brands. We prepare detailed written technical explanations counsel can leverage, and we keep what is known about these systems separate from what is unknown.

  • Learned associations

    Large language models learn statistical patterns from very large bodies of text. People, companies, products, services, and brands are represented largely through those learned patterns, spread across the model, rather than stored as individual profiles.

  • No single record

    There is no single editable record of what a model believes. Some internal representations can be studied, but much of how a model reaches a specific answer remains difficult to observe directly. Change generally comes through the sources a system draws on, adjustments by the provider, and later model versions.

  • Knowledge cutoffs

    Training data ends at a point in time, so a model may repeat information that was once accurate, or know nothing of later events, unless the system retrieves current information.

  • Live retrieval

    Many AI products search for data when a question is asked. An answer can blend retrieved material with what the model learned in training, can shift as those sources change, and does not always show where a particular statement came from.

  • Entity confusion

    Similar names, shared terms, or sparse information can lead a model to merge distinct people, companies, or products, so one party’s history or reputation may be attributed to another.

  • Variation

    Answers can differ across models, versions, prompts, locations, settings, and time, and can change when the same question is asked again.

AI evidence

From a single answer to documented evidence

AI answers can change from one session to the next, and a screenshot shows only that an answer appeared once. We work with counsel to preserve what AI systems said, test how consistently they say it, and explain what the evidence does and does not show. Each finding is documented so it can be reviewed, repeated, and challenged.

  1. 01

    Preserve

    We capture each answer in full, together with the question asked, the platform, the model or version where it is shown, any sources cited, and the date, time, location, and account settings. We also record how each capture was made.

  2. 02

    Reproduce

    We repeat the inquiry across models, versions, wordings, locations, and time, and record where a statement recurs, where it varies, and where it no longer appears. Results are reported with the conditions under which they were observed.

  3. 03

    Explain

    Where the evidence allows, we distinguish answers that appear to draw on training data from those built on live retrieval, identify the sources an answer cites or appears to rely on, and test for confusion with similarly named entities.

Research for counsel

Industry and AI research for legal teams

Legal teams often need the technical facts before they can take a position. We research the AI systems, providers, and industries involved in a matter, test what AI answers say against the documented facts, and present the findings in clear terms. We supply the technical analysis, and counsel provides the legal opinion.

  • How specific AI assistants and models describe a person, business, product, or brand
  • Accuracy testing of AI answers against the documented facts
  • The sources AI answers cite or appear to rely on, and how current those sources are
  • Industry and market research on the AI products, providers, and practices involved in a matter
  • Review of technical documents produced in a matter, including provider documentation and published policies

The Modulus group of companies, founded in 1997, has served clients in more than 100 countries. Modulus research in large language models includes issued U.S. patents on controllable summarization, U.S. 12,008,332 B1 and U.S. 12,380,287 B2, with founder Richard Gardner as a named inventor. That research background informs the analysis we provide to legal teams.[1][2][3]

Working with counsel

Counsel leads the legal work

Modulus is not a law firm and does not provide legal advice. Counsel makes the legal determinations, including whether a matter warrants formal action and how it proceeds. When counsel retains us, counsel directs the work and decides how findings are shared. On request, Modulus can introduce clients to experienced counsel.

Businesses that have not yet engaged counsel can begin with a technical review. Modulus will recommend involving legal counsel when appropriate. Where formal action is warranted, Modulus works with experienced counsel and supports the response with documented technical evidence.

Matters we support

  • False or damaging AI statements about a person, a business, or its products
  • Confusion between similarly named people, companies, or products
  • Warnings or unfavorable recommendations without adequate support
  • Outdated information that persists after the facts have changed
  • A qualified business left out of AI recommendations or ranked poorly in them
  • Technology disputes over what an AI system said or recommended

Modulus AI, Inc. is not a law firm and does not provide legal advice. Counsel makes all legal determinations. Any expert opinion is the independent opinion of the expert, formed from the evidence, and no opinion or outcome is promised in advance.

Bring us the technical question behind your matter.

Counsel and clients can request a confidential consultation to discuss the AI systems, evidence, or research a matter requires.

Request a confidential consultation

Sources

  1. [1]Modulus • Company historyCompany-reported history
  2. [2]U.S. Patent 12,008,332 B1 • Systems for controllable summarization of contentIssued June 11, 2024
  3. [3]U.S. Patent 12,380,287 B2 • Systems for controllable summarization of contentIssued August 5, 2025
Request a confidential consultation