Contents
No matter if we use AI to support or even replace translators, interpreters, or any other profession, there is always that “legal” question lingering in my – and seemingly everybody’s – mind: Is “the AI” being trained on my voice, my translation, my ideas? Will sensitive information end up in this black box LLM and pop up somewhere else inadvertently?
I have been trying to get my head around this question for a while, reading articles and listening to many a legal expert’s explanation. Recently I even had the chance to discuss the matter with the data protection officer of our university (actually I was obliged to do so before buying a CAI tool for my classes). And this, finally, has brought me to the point where I feel enlightened enough to share my most important takeaways for conference interpreters.
Generally speaking, there are different kinds of rights and data that should be treated with caution when dealing with AI – or actually as soon as you enter the digital world.
⚖️ Your and your clients’ rights as an author, speaker, or performer.
These rights were first regulated internationally under the auspices of Victor Hugo! Apart from the Berne Convention from 1886, there is national or EU law depending on your country. These rights include:
- copyright or authors’ rights (depending on the country or legal system you are in): This concerns both your authorship as an interpreter and that of meeting participants.
- neighbouring or related rights to copyright: This concerns the performance as a speaker, again, of both interpreters and speakers.
- moral rights: Moral rights protect the personal relationship between creators and their work, no matter if they own their work or the copyright on it or not. They include the creator’s right to be properly attributed, and the protection of their work from derogatory treatment.
There are also
- personality rights (as part of right over one’s privacy): They concern your and the speakers’ images and voices. Rules seem to be quite different in different countries.
⚖️ Your and your clients’ right to privacy or data protection.
The right to the protection of personal data (a fundamental right according to Article 8(1) of the Charter of Fundamental Rights of the EU): In the EU, according to the famous GDPR, this includes any information relating to an identified or identifiable natural person (‘data subject’) by reference to an identifier such as a name, an identification number, location data, an online identifier or to one or more factors specific to the physical, physiological, genetic, mental, economic, cultural or social identity of that natural person. Privacy or data protection law differs from country to country. Here is a very nice worldwide overview from the University of Berkeley.
⚖️ Your client’s sensitive data.
These are often subject to bilateral NDAs (non-disclosure agreements) between clients and their interpreters. But even if there is no such NDA in place, if we are members of professional associations like AIIC, or VKD in Germany, we are bound to secrecy by the respective codes of professional ethics: AIIC Code of professional ethics, Article 2: Confidentiality: Members of the Association shall be bound by the strictest secrecy, which must be observed towards all persons and with regard to all information disclosed in the course of the practice of the profession at any gathering not open to the public.
Now what does that mean for us as service providers relying on AI-based tools?
As a user (not as a software provider), there is no need for you to know the nitty-gritty of all laws and regulations by heart. You don’t even always have to distinguish between the different kinds of data to be protected. Just make sure that you follow some basic rules:
✔️Read the smallprint of the tools you use.
Before letting a tool “listen” to a meeting and/or your interpretation, or feeding your clients’ texts into a tool, check if, which, and how data is collected, stored, processed, shared and/or used for training. Look for headings like privacy terms, terms of use, data protection, privacy policy. You may need to scroll right to the bottom of the respective page, or click your way through several sub-pages.
✔️Be transparent and ask for permission to feed your clients’ or colleagues’ data into AI-based applications.
Consider which of these data uses pose a risk. Double check with everyone involved what services they are comfortable with. Even if your clients themselves, for example, use WhatsApp for communication, that doesn’t necessarily mean they are happy with you sending files around there.
It might be a professional gesture to publish on our websites which tools and online services we use, and explain in which way and under which conditions. Such tools and services may include generic software like GoogleDrive to share docs and glossaries, Gmail to communicate, DeepL or Generative AI chatbots like ChatGPT, Gemini, Claude, Perplexity, Copilot, or Deepseek, speech-to-text tools like otter.ai or notta.ai, or CAI tools are used for real-time support in the booth.
️✔️Take precautions and follow the principle of data-minimalism.
Neutralise or pseudonimise your text by eliminating names and sensitive data before running it through machine translation.
Use safer alternatives like e-translate, or get paid subscriptions if that saves your data from being stored, processed, shared, and/or being used to train an AI-based system.
Sometimes you can opt-out of your data being used for training models (see, for example, Open AI), or the training even requires your prior consent.
If using Large Language Models, installing them locally may be a good idea in order to protect sensitive data from being shared with third party providers.
I hope this short overview helps to get a first idea of this complex topic. Have I forgotten anything? Do you have suggestions, or best practices to share? Negative experiences, or problems successfully solved? Feel free to share in the comments section below⬇️ or on LinkedIn!
——————-
Disclaimer:
I am not a lawyer and do not pretend to provide legal advice. The above article is the result of extensive reading, listening, and talking to experts, whom I hereby thank fullheartedly. I am also grateful for any clarifications, corrections, additions, or updates. Feel free to leave a comment or contact me bilaterally via email!
Views or opinions expressed are solely my own and do not express the views or opinions of my employer.
————–
About the author:
Anja Rütten has specialised in tech, information and terminology management since the mid-1990s. She holds a professorship in interpreting studies and Computer-Aided Interpreting at the Cologne University of Applied Sciences.
Leave a Reply