With the rise of tools like machine translation engines and generative models (ChatGPT, DeepL, etc.), many companies in the medical and healthcare sector are exploring ways to speed up their multilingual processes. However, when it comes to critical documents —such as IFUs, labelling or SSCPs— speed alone isn’t enough: translations must be accurate, verifiable, and legally compliant.
In this article, we explain what the MDR really says about using machine translation or AI, what risks are involved if these tools are used without professional oversight, and how you can successfully integrate them within a quality system that meets regulatory standards.
Do You Use AI for Translation? Here’s How It Affects MDR Compliance
In the medical device sector, translation is not optional: it is a legal obligation. Regulation (EU) 2017/745 (MDR) requires that all documentation — from Instructions for Use (IFUs) to labelling and the Summary of Safety and Clinical Performance (SSCP) — must be available in the official language of every country where the product is marketed.
Translation is therefore not a mere administrative step. It is a regulatory requirement directly linked to patient safety and the correct use of the device. A single mistranslated word, concept or instruction may lead to misuse, adverse events or, ultimately, regulatory sanctions for the manufacturer.
With the arrival of generative artificial intelligence and large language models (LLMs), many companies are asking themselves the following:
Can we speed up translations with AI without jeopardising regulatory compliance?
The answer is clear: yes, but only within a controlled process and with certified professional review.
Machine Translation or Generative AI? Both Require Human Review
Both machine translation engines and AI models (such as ChatGPT) can generate multilingual content quickly and fluently. However, in the medical field, speed without control can have serious consequences.
Common Terminology Pitfalls
Certain terms and expressions are commonly misinterpreted by machine translation systems when translating for instance from Spanish into English in medical and regulatory contexts. These are real examples where AI or automatic translation tools frequently produce incorrect, imprecise, or misleading results without human post-editing.
“Dispositivo médico” → Machine translations often render this too generically.
✅ Correct: Medical device (MDR terminology).
❌ Incorrect: Appliance, instrument, tool, implement.
“Producto sanitario” → “Sanitario” is tricky. MT engines may translate this as hygiene product (e.g. pads, soap).
✅ Correct: Medical product or healthcare product (depending on context).
❌ Incorrect: Sanitary product (especially in UK English, where it refers to hygiene/personal care).
“Marcado CE” → MT often struggles with this formulation and over-literal variants.
✅ Correct: CE marking
❌ Incorrect: CE mark, stamp, label
“Uso previsto” → MT may render this literally as expected use or planned use.
✅ Correct: Intended use (as defined in MDR Article 2)
❌ Incorrect: Expected use, foreseen use, anticipated use
“Datos clínicos” → Too many variations can be introduced incorrectly.
✅ Correct: Clinical data (MDR Annex XIV)
❌ Risky: Clinical facts, medical information (too vague)
And so on…
Using AI without expert linguistic supervision risks introducing exactly these kinds of errors. Misinterpretations can compromise patient safety and place product compliance in jeopardy. And without documented traceability or quality control, any audit becomes an additional liability.
Is It Feasible to Manage Translations with AI In-House?
Yes, it is possible — but only under strict conditions.
If a company has qualified staff in medical translation, applies documented processes of professional review, and ensures full traceability, it may consider in-house management. This model can be suitable for less critical content, such as marketing materials or internal communications.
However, for sensitive documents — IFUs, labelling, clinical or technical documentation — the risk of non-compliance is high without proper professional review. A mistranslated IFU not only breaches the MDR but may also trigger product withdrawal from the market.
In practice, most manufacturers prefer to rely on specialised external providers, since maintaining a complete linguistic quality system internally is costly and difficult to sustain in the long term.
ISO 17100 vs ISO 18587: Which Standard Applies?
Two key standards ensure quality in language services:
🔵 ISO 17100 – Human translation only. Requires at least two native specialist linguists (one translates, another reviews), strict terminology control and rigorous quality processes.
🟡 ISO 18587 – Post-editing of machine or AI translation. Defines how automatically generated content must be reviewed to meet professional standards and avoid risks to compliance or safety.
If your company uses automated tools, ISO 18587 is the standard that should guide the process.
A key point: the MDR does not explicitly mention ISO 17100 as a mandatory requirement. However, it does require that all information be understandable and reliable for the end user. Therefore, AI or machine translation can be used, provided the following conditions are met: professional review, traceability and guaranteed quality.
How Do Authorities Audit Linguistic Compliance?
During inspections and audits, Notified Bodies typically verify that:
- All mandatory documents are available in the required languages.
- A documented process of translation and review exists, with identified responsible parties.
- Full traceability is maintained (who translated, who reviewed, which tools were used, what changes were applied).
- Terminology control and linguistic validation measures can be demonstrated.
If inconsistencies, incomplete translations or lack of professional review are detected, the outcome may be a major audit finding or even suspension of CE conformity.
What Does a Provider Certified in ISO 17100 and ISO 18587 Offer?
Working with a certified partner not only reduces risk but also frees up internal resources. Key benefits include:
✔️ Professional review by linguists specialised in the medical field.
✔️ Linguistic and technical quality control at every stage.
✔️ Full process traceability (who translates, who reviews, which tools are used).
✔️ Compliance with MDR and other international regulatory frameworks.
✔️ Legal security in audits and inspections.
✔️ Long-term cost optimisation through translation memories and centralised terminology management.
Conclusion: Efficiency Yes, But Never at the Expense of Quality
AI can accelerate processes and optimise resources. But in the medical device sector, efficiency without quality is not an option.
If your organisation wishes to benefit from new technologies in translation, make sure to do so with full safeguards. Collaborating with a provider certified in ISO 18587 allows you to combine the agility of automation with the quality control needed to comply with the MDR.
Ultimately: using AI for translation does not mean breaching the MDR — provided there is a professional process to support it.
Therefore, if you are going to use AI or machine translation, do it properly: with professional review and under a recognised quality management system.
Want to ensure your translations are fully MDR-compliant?
Contact our team and discover how we integrate AI within an ISO 18587 certified process.
