Why Whisper Can’t Replace Court Reporters in the U.S. Legal System

Posted bystenoimperiumPosted inAI and the Legal System, Court ReportingTags:, , , , , , , , , , , , , , , , , , , , , ,

In a world racing to adopt artificial intelligence, it’s tempting to view new speech-to-text tools like OpenAI’s Whisper as the holy grail of courtroom transcription. Whisper, an open-source ASR (automated speech recognition) model trained on over 680,000 hours of audio, has received widespread praise for its accuracy and flexibility. But before the U.S. court system follows the siren call of automation, it would do well to examine how Canada’s Legislative Assembly of British Columbia approached its adoption of this tool — and what lessons the U.S. legal system should heed.

A recent article in TIRO: The Journal of Professional Reporting and Transcription, entitled “Harnessing Whisper at the Legislative Assembly of British Columbia: A User-Driven Approach to AI-Supported Parliamentary Reporting” details how British Columbia implemented Whisper into its parliamentary reporting workflow. They did not use it to replace human reporters or editors. Instead, they created a hybrid system called Parrot, where ASR supports — but does not replace — trained editorial staff. The results? Faster first drafts, enhanced productivity, and a strong affirmation of the critical role humans play in maintaining accurate, reliable records.

It’s a cautionary tale for the United States, where some courts, agencies, and law firms are aggressively replacing stenographic reporters with ASR and digital recorders. In doing so, they risk sacrificing the integrity of the legal record — and ultimately, the constitutional rights of the people the system is meant to serve.

The Illusion of Accuracy

OpenAI’s own benchmark reports suggest Whisper achieves a word error rate (WER) of around 8.81% in general settings. With human oversight — what’s known as a “human-in-the-loop” approach — that can drop to 7.61%. But even that level of error, applied across hours of legal proceedings, results in thousands of inaccuracies. In a courtroom, where every word matters, every syllable can shift the balance of justice.

By contrast, certified court reporters in the U.S. routinely achieve accuracy rates of 99% or higher. We are trained not only to capture verbatim speech but also to manage cross-talk, regional dialects, technical jargon, and fragmented or interrupted statements — the kinds of nuanced speech that ASR still struggles to comprehend. In fact, ASR models like Whisper tend to fail most spectacularly in exactly the situations where court reporters excel: emotionally charged testimony, overlapping speakers, and highly specialized terminology.

Speaker Attribution Is Still a Major Obstacle

One of the most glaring weaknesses of current ASR tools — including Whisper — is speaker diarization, the ability to accurately identify and label who is speaking. In the British Columbia implementation, this issue was so critical that they had to integrate microphone data from their sound system to correctly attribute speech to the right member of parliament.

Now imagine that problem in a U.S. courtroom, deposition, or hearing. Speaker misattribution doesn’t just cause confusion — it can result in mistrials, sanctions, and appeals. Attribution is not optional. It’s essential. And any system that lacks robust speaker ID is fundamentally unfit for legal use.

Data Privacy and Security

British Columbia also made the wise decision to implement Whisper locally, rather than through cloud servers, in order to protect sensitive data. The U.S. court system, however, is seeing a rise in the use of third-party vendors who upload deposition audio and courtroom recordings to offsite servers — often without the knowledge or informed consent of litigants, attorneys, or even the court.

This practice raises profound legal and ethical concerns. Privileged conversations, sealed testimony, and sensitive criminal or civil proceedings should never be entrusted to cloud servers outside the jurisdiction of the court. In an age of increasing data breaches and cyberattacks, local control isn’t just preferable — it’s essential.

The Myth of Cost Savings

One of the primary drivers behind the adoption of ASR in the U.S. legal system is cost. But this is a false economy. What’s saved in labor is lost in post-processing, appeal litigation, and the credibility of the record. AI-generated transcripts often require extensive correction by attorneys or court staff, shifting the burden — and the cost — onto those already stretched thin.

British Columbia’s approach is a model of responsible integration. They used ASR to free up human editors for higher-level tasks like formatting, research, and verification — not to eliminate their roles entirely. They preserved editorial control and ensured that the final product remained human-reviewed and human-approved.

This distinction is critical: When ASR is implemented with humans, it can improve efficiency. When it’s implemented instead of humans, it compromises accuracy, fairness, and the constitutional right to a complete and correct record.

Due Process Demands a Human Record

The U.S. Constitution guarantees the right to due process, which includes the right to an accurate and complete legal record. This principle is upheld through case law, ethical standards, and procedural rules. If someone is wrongly convicted because a computer failed to capture a key statement, or if a deposition is misinterpreted because an AI mistranscribed medical terminology, who is held accountable?

ASR cannot swear an oath. It cannot be cross-examined. It cannot be held liable. A court reporter can — and does — stand behind the integrity of their record. That is a safeguard worth preserving.

Let Professionals Lead the Way

The most striking thing about the TIRO article is that the professionals implementing the technology were not technologists — they were editors, reporters, and transcription experts. Their deep domain expertise guided every decision. They didn’t blindly adopt AI; they shaped it to serve their standards, not the other way around.

This is the path forward. Rather than letting software vendors dictate how legal records are made, the U.S. court system must empower its stenographic professionals to lead the adoption of supportive technologies. Court reporters are not anti-tech. We are pro-accountability. Pro-accuracy. Pro-justice.

Technology Must Serve Justice, Not Undermine It

Whisper may be an exciting tool. But it is just that — a tool. It is not a replacement for human court reporters, especially in adversarial legal environments where every word counts, every speaker matters, and the consequences of error are profound.

British Columbia got it right. They harnessed Whisper to support their professional team — not to replace them. The U.S. legal system must take note, before it’s too late. We need human-in-the-loop systems, not human-out-of-the-courtroom experiments.

Justice deserves better than a rough draft. It deserves the truth — accurately captured, ethically safeguarded, and professionally delivered.

StenoImperium
Court Reporting. Unfiltered. Unafraid.

Disclaimer

The content of this post is intended for informational and discussion purposes only. All opinions expressed herein are those of the author and are based on publicly available information, industry standards, and good-faith concerns about nonprofit governance and professional ethics. No part of this article is intended to defame, accuse, or misrepresent any individual or organization. Readers are encouraged to verify facts independently and to engage constructively in dialogue about leadership, transparency, and accountability in the court reporting profession.

  • The content on this blog represents the personal opinions, observations, and commentary of the author. It is intended for editorial and journalistic purposes and is protected under the First Amendment of the United States Constitution.
  • Nothing here constitutes legal advice. Readers are encouraged to review the facts and form independent conclusions.

***To unsubscribe, just smash that UNSUBSCRIBE button below — yes, the one that’s universally glued to the bottom of every newsletter ever created. It’s basically the “Exit” sign of the email world. You can’t miss it. It looks like this (brace yourself for the excitement):

Published by stenoimperium

We exist to facilitate the fortifying of the Stenography profession and ensure its survival for the next hundred years! As court reporters, we've handed the relationship role with our customers, or attorneys, over to the agencies and their sales reps.  This has done a lot of damage to our industry.  It has taken away our ability to have those relationships, the ability to be humanized and valued.  We've become a replaceable commodity. Merely saying we are the “Gold Standard” tells them that we’re the best, but there are alternatives.  Who we are though, is much, much more powerful than that!  We are the Responsible Charge.  “Responsible Charge” means responsibility for the direction, control, supervision, and possession of stenographic & transcription work, as the case may be, to assure that the work product has been critically examined and evaluated for compliance with appropriate professional standards by a licensee in the profession, and by sealing and signing the documents, the professional stenographer accepts responsibility for the stenographic or transcription work, respectively, represented by the documents and that applicable stenographic and professional standards have been met.  This designation exists in other professions, such as engineering, land surveying, public water works, landscape architects, land surveyors, fire preventionists, geologists, architects, and more.  In the case of professional engineers, the engineering association adopted a Responsible Charge position statement that says, “A professional engineer is only considered to be in responsible charge of an engineering work if the professional engineer makes independent professional decisions regarding the engineering work without requiring instruction or approval from another authority and maintains control over those decisions by the professional engineer’s physical presence at the location where the engineering work is performed or by electronic communication with the individual executing the engineering work.” If we were to adopt a Responsible Charge position statement for our industry, we could start with a draft that looks something like this: "A professional court reporter, or stenographer, is only considered to be in responsible charge of court reporting work if the professional court reporter makes independent professional decisions regarding the court reporting work without requiring instruction or approval from another authority and maintains control over those decisions by the professional court reporter’s physical presence at the location where the court reporting work is performed or by electronic communication with the individual executing the court reporting work.” Shared purpose The cornerstone of a strategic narrative is a shared purpose. This shared purpose is the outcome that you and your customer are working toward together. It’s more than a value proposition of what you deliver to them. Or a mission of what you do for the world. It’s the journey that you are on with them. By having a shared purpose, the relationship shifts from consumer to co-creator. In court reporting, our mission is “to bring justice to every litigant in the U.S.”  That purpose is shared by all involved in the litigation process – judges, attorneys, everyone.  Who we are is the Responsible Charge.  How we do that is by Protecting the Record.

2 thoughts on “Why Whisper Can’t Replace Court Reporters in the U.S. Legal System”

  1. Am I correct in understanding that Canada does not use court reporters? Is everything handled by editors and proofreaders?

    Like

    Reply

    1. Great question! You’re partly correct — but there’s important context. In some Canadian provinces, like British Columbia, parliamentary reporting and certain court or legislative transcription functions are handled by editorial teams who work from high-quality audio recordings, often supported by ASR (automated speech recognition) like Whisper.

      However, Canada’s court systems vary by province. In formal judicial proceedings (especially in higher courts), stenographic court reporters are still used, particularly in complex trials or appeals. Other courts may rely on digital recording systems, with transcripts produced post-proceeding by transcriptionists or editors, not real-time stenographers. This differs from many parts of the U.S., where a live court reporter is typically present to capture the record in real time, often with certification requirements.

      The British Columbia example in the article is about parliamentary transcription — not criminal or civil courts — and shows how they’ve responsibly implemented AI to support, not replace, professional staff.

      Like

      Reply