When the Record Speaks — and Software Interprets – The Unsettled Ethics of AI Deposition Summaries

In courtrooms and conference rooms across the country, a quiet transformation is taking place in the handling of sworn testimony. For generations, the deposition transcript stood alone: a verbatim record created by a neutral officer whose duty was limited to capturing words exactly as they were spoken. The record contained no interpretation, no commentary, and no conclusion. Its authority rested on that restraint.

Today, the transcript is no longer the final product. Increasingly, court reporting agencies now offer an additional service: automated summaries generated by artificial intelligence and sold alongside the official record. Attorneys may purchase condensed narratives explaining what witnesses said, often within minutes of receiving the transcript. The summaries promise efficiency and convenience, and in a profession shaped by deadlines, both are welcomed.

But the practice has introduced a new and unsettled question into the legal system. When a sworn record becomes the source of a commercial interpretation, does the neutrality of the reporter who created it remain intact?

The debate has centered on a decades-old ethics guidance, Advisory Opinion 32 of the National Court Reporters Association. The opinion predates machine learning and litigation platforms by decades, yet it has become the focal point of an argument about whether technology has altered the boundaries of the reporter’s role or merely obscured them.

The answer depends less on what software does than on what the legal system believes the reporter has done.

The Record, the Algorithm, and the Appearance of Neutrality

Advisory Opinion 32 was written to address a simple situation. The ethics committee considered whether a court reporter could prepare a summary of testimony taken in a deposition. The committee concluded the reporter should not. The reason was not accuracy, competence, or even bias. It was perception. If the same neutral officer responsible for preserving testimony also produced an explanation of it, a litigant could reasonably believe the reporter exercised judgment about the meaning of the evidence. Even if the summary were written by another employee of the same reporting firm, the association could not be eliminated. The public might still assume the reporter contributed to the interpretation.

The opinion therefore protected not only neutrality but the appearance of neutrality. The reporter’s function was to capture words, not to explain them.

Artificial intelligence has complicated that principle by inserting a new actor between record and reader. The modern workflow often proceeds as follows: the reporter captures testimony, the agency distributes the transcript, software analyzes the text, and the agency sells the resulting summary. The reporter neither requests nor reviews the analysis. In many cases the reporter does not know it exists. Because no human connected to the reporter wrote the summary, some have concluded the ethical concern disappears.

Neutrality for Sale – When Platform Logic Rewrites Court Reporter Ethics

Among those presenting that conclusion publicly is industry commentator Lindsey Stoker, who has argued that automated summaries fall outside the scope of the ethics opinion because software is not a certified reporter and therefore cannot implicate the reporter’s professional duties. The assertion has been framed as settled interpretation, offered without judicial ruling, disciplinary precedent, or regulatory guidance confirming it.

The confidence of the conclusion exceeds its legal foundation.

Stoker is also affiliated with Filevine, a litigation technology platform whose ecosystem benefits from the continued availability of automated transcript-analysis tools. That relationship does not invalidate her reasoning, but it provides context: the interpretation she advances aligns with a commercial model that depends on third-party software vendors generating analytic products from deposition testimony. In professional ethics debates, positions that intersect with active business interests typically warrant heightened scrutiny, particularly where the conclusion would normalize a revenue-producing service whose regulatory status remains unresolved.

Ethics rules governing neutral officers have rarely depended on authorship alone. Courts regularly examine whether conduct creates the appearance of influence, not merely whether influence occurred. Judges recuse themselves for potential bias, not proven bias. Attorneys disclose conflicts that may never materialize. The law recognizes that trust in process is as important as the process itself.

Advisory Opinion 32 follows that tradition. The committee did not ask who physically wrote the summary. It asked whether an observer could reasonably believe the reporter was connected to an interpretation of testimony. Artificial intelligence does not eliminate that perception. In certain business models, it may reinforce it.

When summaries are marketed through the same portal that delivers the transcript, when they derive exclusively from the official record, and when they are sold by the same provider that arranged the reporting services, the distinction between record and explanation becomes invisible to the user. From the attorney’s perspective, the summary appears to carry the authority of the record itself. The consumer of the service is not expected to parse corporate structures separating reporter, agency, platform, and vendor. Ethics rules exist precisely because participants in litigation cannot investigate those relationships before relying on the record.

The claim that the ethics concern disappears because software authored the summary treats the rule as if it regulated typing rather than trust. It converts a principle designed to protect confidence in neutrality into a technical question about who pressed the keys.

The practical consequences extend beyond theory. If an automated summary mischaracterizes testimony and influences litigation strategy, the dispute ultimately returns to the transcript. Attached to that transcript is the sworn officer who created it. The software company is not called to authenticate testimony. The reporter may be asked to clarify what the witness actually said, provide declarations, or respond to challenges arising from the interpretation. Even if the reporter bears no fault, the reporter becomes entangled in a disagreement over meaning, precisely the circumstance the ethics opinion sought to prevent.

This exposure has prompted some reporters to notify agencies they do not authorize derivative interpretive products generated from their transcripts. The notices do not attempt to stop the summaries. They serve to document separation, establishing that the reporter neither prepared nor endorsed any explanation of testimony. The need for such disclaimers illustrates that the boundary between record and interpretation is no longer obvious.

Economic incentives further complicate the matter. Deposition summaries were historically attorney work product prepared by lawyers and billed hourly. Automation has converted interpretation into a purchasable commodity. Reporting agencies now monetize not only the capture of testimony, but its analysis. The reporter’s transcript provides the foundation for the product, yet the reporter exercises no control over how testimony is characterized while remaining attached to the authority that gives the characterization value.

This shift has produced a structural gap between professional duty and commercial practice. The reporter remains the neutral officer responsible for the record’s credibility, while the meaning of that record becomes a marketable add-on.

Stoker’s position, that automated summaries categorically fall outside the ethics rule, attempts to resolve the gap by redefining the rule’s reach rather than examining its purpose. The reasoning depends on the absence of precedent as proof of safety. Yet the absence of precedent in emerging technology rarely establishes certainty. It usually marks the point before adjudication occurs.

To present a contested interpretation as authoritative risks more than disagreement. It risks reshaping professional expectations before courts or regulators have addressed whether attribution, not authorship, governs the reporter’s neutrality in the era of automated analysis.

The issue ultimately turns on a broader question: whether testimony is treated as data or as sworn evidence requiring human stewardship. If testimony is merely information, automated interpretation presents little concern. If testimony carries the weight the legal system has historically assigned it, then any commercial explanation attached to the official record must be unmistakably separate from the authority of the officer who created it.

The legal system has long chosen the latter view. Evidence differs from information because it influences rights, obligations, and outcomes. The reporter’s oath exists to preserve confidence that what is recorded reflects only what was said. The more closely interpretation travels with the record, the more that confidence depends on the clarity of their separation.

Artificial intelligence has not changed the reporter’s duty. It has changed the environment surrounding it. Words once preserved and then interpreted by advocates can now be interpreted automatically and sold beside the record itself. The technology functions efficiently. The ethical boundary, however, remains unsettled.

Courts, bar associations, or regulatory bodies will eventually decide whether automated interpretive products attached to the official transcript alter the perception of neutrality. Until then, assertions of certainty are premature. The question has not been resolved; it has merely been postponed by convenience.

The reporter’s role has always rested on restraint. The authority of the transcript comes from the understanding that the record speaks for itself. As technology grows capable of explaining testimony instantly, the legal system must determine whether that explanation travels independently or under the shadow of the oath.

The distinction will decide whether the reporter remains the silent custodian of words or becomes, however unintentionally, connected to their interpretation.

---

Disclaimer

This article is commentary and analysis for informational purposes only and does not constitute legal advice or an ethics ruling. Readers should consult applicable statutes, court rules, and professional licensing authorities in their jurisdiction for guidance regarding specific practices.