FCA authorised · FRN 724952 0117 325 0027Quote & buy →
Aipedia: Translators and interpreters
Feeding a client's translation memory or glossary into an AI engine can breach confidentiality or contract terms if the client did not agree to it. Who owns that linguistic data, and how it may be reused, should be settled in writing.
Part of: AI in Translators and interpreters
In short
Translation memories (TMs) and termbases hold years of a client's translated content and approved terminology. Translators and agencies may use them to customise MT engines or prompt LLMs, which can improve consistency. But TMs often contain confidential and personal data, and clients may own them under contract. Uploading them to a public tool, reusing one client's TM for another, or using them to train an engine without permission can lead to confidentiality, data protection and IP disputes. PI may respond to some of these claims and cyber cover to others, depending on the policies.
CIOL's AI Voices white paper cites a survey of Association of Translation Companies members: 82% used translation memory and 71% used MT, against 25% using LLMs or generative AI. The tools are converging. Agencies use client TMs to adapt MT engines, translators paste glossary extracts into prompts to steer a generative AI chatbot, and some platforms offer to "learn" from everything a user uploads.
A TM is not just a list of sentences. It may contain names, contract terms, unreleased product details, medical information and anything else the client ever sent for translation. A glossary may encode a client's brand and legal terminology built up over years.
That is why clients may ask suppliers how linguistic assets are stored and whether any tool provider can see or keep them. A translator who cannot answer those questions clearly may lose work, and one who gives a confident answer that turns out to be wrong may face a claim.
Your client contract, any NDA and data protection law set the rules. If you process personal data for a client, you are usually a processor and need Article 28 terms under UK GDPR; using that data in a tool the client did not approve may fall outside your instructions. The ICO's AI guidance applies to tools that process personal data. CIOL warns users to know what data they give AI tools and where it goes. Ownership of a TM is usually a contract question; many agency and client terms say the client owns it.
Illustrative scenario: a freelancer uses a large TM from a pharmaceutical client to customise an online MT engine, so as to work faster on the client's next project. The tool's terms allow the provider to use uploaded content to improve its service. The client later learns of this during a supplier audit, says confidential trial information and personal data were disclosed in breach of contract and its NDA, and claims the costs of investigating and notifying.
Other disputes in this area include:
A claim that you negligently disclosed confidential client information, or breached confidentiality in the course of professional work, may be covered by translators' PI, subject to the wording. Many PI policies include breach of confidence, while data breach costs such as investigation and notification are more often a cyber policy matter. Claims based on breach of a contract term you agreed, rather than negligence, may be treated differently. Our page on AI-generated content and IP infringement looks at how PI and cyber split these risks.
To keep linguistic data safe when using AI, put these steps in place:
Related guides: client confidentiality and AI tools, machine translation post-editing and AI transcreation and marketing copy.
Apex arranges PI for translators and agencies, and can look at cyber cover alongside it. A broker checks how your policies treat confidentiality and data claims arising from AI tools; see cyber insurance and AI risks. Return to the AI in translation and interpreting hub.
It depends on the contract. Many client and agency terms say the client owns the TM and glossaries created for it. Where nothing is agreed, ownership can be disputed. Settle it in writing, including what you may do with the data afterwards.
Only if the client agrees and the tool's terms protect the data. A TM can contain confidential and personal data, and uploading it to a tool that reuses content may breach your contract, an NDA and data protection law.
Possibly both. Translators' PI often covers claims for breach of confidence in professional work, while costs such as investigation and notification after a data breach are more usually cyber matters. Check both wordings with your broker.
Tell us about your firm and the work you do. A broker will review how your current PI treats AI-assisted work, or arrange new cover. No fee for a quote. Or call 0117 325 0027.
Start a PI quote Request a callbackApex Insurance Brokers Limited is authorised and regulated by the Financial Conduct Authority. Registered in England and Wales, company number 07014570. This page is general information about how AI is used in the profession and how insurance may respond. It is not legal, regulatory or insurance advice on your circumstances, and it does not guarantee that cover will be available or on what terms.
