Using AI Transcription for Court Proceedings? A Dark Day In the Life of a Translator...

AI Transcription: A Dystopian Future?

I know, I know... Not ANOTHER post about AI and how it's coming for our jobs!

While I'm normally quite "upbeat" about embracing AI, one particular interaction with a translation agency left me feeling compelled to share my thoughts on the current state of the industry.

I've worked as a translator for 15 years, and it's no exaggeration to say that this marked a "dark day" in my career...

So, in this blog post, I want to share what happened, and share some insights and advice for colleagues, agencies and end clients alike.

The Context

Not long ago, I was contacted by an agency (just to add: this is a legitimate agency I've worked with for years) about a job that seemed straightforward at first glance: edit the transcription and translation of some Cantonese voice recordings into English for use in court proceedings.

But as soon as I reviewed the files, I knew something was very wrong.

The recordings were incredibly complex – multiple people talking over each other, voices raised in heated argument, curse words flying, and a constant mix of abbreviations and slang. It was the kind of speech that even a seasoned human transcriber has to pause, replay, and really think carefully about.

And the transcription I was given to "edit" was, to put it mildly, a mess.

Whole phrases were missing.

Abbreviations were mangled.

Swear words and slang were either misinterpreted or left out entirely.

There was NO WAY this transcription had been performed by a human translator, even the most incompetent translator.

Which can only mean one thing.

It had been done by AI!

When “Editing” Really Means Starting From Scratch

The agency wanted me to “edit” the transcripts, but it quickly became clear this wasn’t an editing job at all.

If these files were to stand up in court, they needed a full human transcription and translation – done from scratch.

I explained this to the agency and gave them a realistic quote for the work. But they came back to say the budget simply "couldn’t stretch that far".

What happened next, I don’t know – whether the case went ahead with a cheaper option, or whether the end client abandoned it altogether.

But the incident left me... shaken.

A Sobering Moment

Look: I'm far from an AI "basher"!

I’ve generally tried to stay positive and upbeat about the impact of AI on our profession. I see its potential as a tool, and I believe there are ways it can support translators and linguists. Even in the realm of transcription, I think AI can be extremely valuable for some use cases, such as transcribing meeting notes and videos.

But this particular incident hit me hard.

Have we really reached the point where agencies believe it’s acceptable to submit low-quality AI transcripts and translations for court cases – cases that could decide a family’s future forever?

That thought made me feel, at least for a moment, deeply sceptical.

Pessimistic.

Like we were staring at a dystopian future for our profession.

Why AI Isn’t There Yet

As I mentioned earlier, AI transcription tools definitely have their place, but they are FAR FROM PERFECT. Here are some key reasons why AI tools cannot be relied on for high-stakes content:

• Multiple speakers: AI can’t handle people talking over one another.

• Slang and abbreviations: These get misinterpreted or ignored.

• Context and culture: Machines don’t pick up tone, implication, or cultural nuance.

• Accent: AI cannot possibly understand the myriad accents across all languages!

• Minority languages: AI tools are mainly trained on the world’s most commonly used languages, so minority languages such as Cantonese are often underrepresented – making results far less accurate.

• Accountability: AI tools cannot be accountable. Only a human translator with a real name and professional credentials can ever provide accountability!

Practical Advice for Agencies and Clients

This experience also made me reflect on what agencies, end clients, and even fellow linguists can do to prevent situations like this. Here are a few practical takeaways:

• Be transparent from the start.

  If AI has been used in the workflow, let the linguist know upfront. Don’t frame it as a light “editing” job if the output is unusable – it only wastes time for everyone involved. And PLEASE be transparent with the end client, too, if AI is used in any step of the process!

• Set budgets realistically.

High-stakes transcription and translation isn’t a quick or cheap task. It requires concentration, cultural knowledge, and professional judgment. If the outcome could affect people’s lives, underfunding the work is a false economy, not to mention inethical!

• Choose qualified linguists.

For sensitive or legal cases, work with certified professionals who can produce work that stands up in court.

• Use AI wisely.

AI can be great for low-risk, internal purposes like meeting notes or rough drafts. But for "messy" recordings, minority languages, or legal proceedings, AI should never be relied upon as the final product; in fact, I argue that as things stand now, there's no point using it at all as it just adds an unnecessary step to the process.

Standing Our Ground

So where does this leave us as language professionals?

For me, it’s a reminder of why our role as professional linguists is more important than ever.

We have to stand our ground and make the case, again and again, that human expertise is not optional in high-stakes contexts.

Maybe one day AI will reach a point where it can handle messy, emotional, multi-speaker recordings with complete accuracy.

But even then, human oversight will remain 100% essential.

Why? Because high-stakes translations have high-stakes consequences.

They can have ripple effects on the lives and futures of individuals – and entire families.

And that’s something no machine can ever fully carry on its shoulders.