AI is making its way into government language services, but a recent parliamentary answer suggests that accuracy, accountability and human expertise remain firmly on the agenda. Clare Suttie, Director of Atlas Translations, explains why this matters well beyond the justice system.
Even the Ministry of Justice is asking the right questions about AI translation. Are you?
Hardly a day goes by without another headline about artificial intelligence replacing somebody’s job. Translators, interpreters, lawyers, accountants, journalists. And yet, when you look beyond the headlines, the picture is rather more interesting.
On 2 November 2026, the Ministry of Justice (MoJ) published a response to a parliamentary question about its spending on translation and interpreting services, and its plans to introduce more AI into those services. Given that I’ve spent more than 35 years working in the language industry, it certainly caught my attention. Not because the MoJ is looking at AI. That’s hardly surprising.
What interested me was how carefully it appears to be approaching the question of where AI is appropriate, and what safeguards need to be in place. And I think there’s a lesson here for every organisation that buys translation or interpreting services.
What exactly has the Ministry of Justice said?
In response to a question from Lord Spellar, the MoJ confirmed that it is exploring AI-assisted interpreting for non-legal conversations in prisons, alongside AI-assisted translation of publications.
Crucially, the latter includes:
“human quality assurance of translated material.”
That might sound like a small detail, but it’s an important one.
The MoJ also confirmed that its new language services contracts, which came into effect on 5 October 2026, include requirements for suppliers to monitor and report on their use of AI and innovation, including the impact on interpreter availability and accuracy. And before any wider use of AI is considered, the department says it will evaluate accuracy, consistency, suitability across different languages and settings, data protection, security and human oversight. In other words, it’s not simply asking whether AI can translate or interpret something. It’s asking whether it can do so reliably, in the circumstances in which it will be used. That’s a rather different question.
Because getting the words right isn’t always enough
At Atlas Translations, we work with organisations in sectors where accuracy really matters: healthcare, pharmaceuticals, legal services, manufacturing, charities and many others. And one thing I’ve learnt over the years is that two documents of exactly the same length can carry completely different levels of risk.
A translated internal newsletter might contain a slightly awkward phrase. Annoying, perhaps, but probably not disastrous.
A mistranslated medical instruction, legal agreement or safety warning? That’s another matter entirely.
The consequences could be serious.
AI can produce a translation that reads beautifully and is still wrong.
That’s one of the challenges with relying on automated translation. A fluent sentence can inspire confidence, even when it has changed the meaning, omitted a qualification or misunderstood the context. A reader who doesn’t speak the original language may have absolutely no way of knowing.
Which is why I think the conversation about AI and translation needs to move on. We’ve spent far too long asking whether AI is better, faster or cheaper than a human translator. I’d much rather ask: What happens if this translation is wrong?
The human element isn’t an optional extra
The MoJ’s announcement comes against the backdrop of new language services contracts designed to improve consistency, reliability and accessibility across courts and tribunals. In its announcement on 5 October, the government explained that the new arrangements include strengthened interpreter induction, professional standards and independent quality assurance.
It described the objective as:
“fair and effective access to justice.”
And that is really what this is all about.
Language services aren’t simply about converting words from one language into another. They’re about making sure people understand what’s happening, can make informed decisions and have a fair opportunity to participate. Imagine being involved in a legal proceeding in a language you don’t understand. Or trying to follow medical advice in a language you barely speak. Would you want the cheapest possible solution? Or would you want to know that someone appropriately qualified was responsible for getting it right? I know which I’d choose.
Of course, professional translators and interpreters aren’t infallible either. That’s precisely why qualifications, experience, professional judgement and appropriate quality assurance processes matter. Human involvement shouldn’t be a box-ticking exercise at the end of an automated process. It needs to be meaningful, properly resourced and appropriate to the risks involved.
We’re not anti-technology. We’re pro-getting-it-right.
I should probably make something clear at this point. Atlas Translations isn’t against technology. Far from it. We’ve been working with translation technology for decades, and our industry has continually adapted as new tools have emerged. We also offer machine translation post-editing (MTPE) as a separately commissioned service where it’s suitable for the material and the client’s requirements. But we don’t believe that every document should be pushed through an AI translation tool simply because it’s available. Nor do we think that having a human glance over the output automatically makes it safe or suitable for every purpose.
Sometimes professional human translation is the right answer. Sometimes a carefully managed machine translation post-editing workflow may be appropriate. Sometimes the first thing a client needs isn’t a translation at all, but a conversation about what they’re trying to achieve, who needs to understand the material and how it will be used. That’s the part of our job I find particularly interesting. After 35 years, I still think one of the most valuable things we can do is help a client work out what they actually need, rather than simply provide what they’ve asked for.
Three questions every organisation should ask
If you’re responsible for buying translation or interpreting services, I’d suggest starting with three questions.
1. What happens if the meaning is wrong?
Is the worst outcome a slightly confusing sentence, or could somebody’s health, safety, legal position or reputation be affected?
2. What will people use this information for?
There’s a considerable difference between content intended for general background reading and something that will inform a medical decision, form part of a contract or be relied upon in legal proceedings.
3. How will we know the work is accurate?
What checks are needed? Who is qualified to carry them out? And what records should be available afterwards?
These aren’t questions that only apply to AI. They’re questions about quality, responsibility and making sensible decisions. And they should be part of every serious conversation about language services.
What happens next?
The MoJ’s parliamentary answer included some interesting figures. During 2025, it recorded approximately £30.1 million in contract expenditure on spoken-language services and almost £3 million on translation and transcription. These are substantial sums, although it’s important to recognise that they refer to contract expenditure, n ether technology can improve services, increase availability or deliver better value.
But value for money and the lowest possible price aren’t necessarily the same thing. The real test will be what the MoJ’s evaluations reveal, how transparently the results are reported and whether any expansion of AI is supported by evidence rather than enthusiasm alone.I’ll be watching with interest.
The future of translation isn’t simply human versus machine
I suspect we’ll continue to see plenty of dramatic predictions about the end of translation as we know it. Some will make interesting reading. Others will make those of us working in the industry roll our eyes over our morning tea. But I do think the more useful conversation is already beginning.
It’s not about choosing sides in an AI debate. It’s about understanding risk, selecting the right workflow and making sure that the people who rely on translated information can trust it. If the Ministry of Justice is taking the time to examine accuracy, security, suitability and human oversight before expanding its use of AI, perhaps other organisations should be asking themselves similar questions.
Because the most important question isn’t whether AI can translate something.
It’s whether the translation can be trusted when it matters.
And that’s where professional language expertise remains invaluable.
Clare Suttie is Director of Atlas Translations, an award-winning, ISO-certified translation and interpreting company established in 1991. Atlas helps organisations communicate accurately and effectively across languages.



