AI transcription is transforming academic research, but researchers uploading confidential interviews to online services could be putting their participants’ personal information, research ethics approval and university data protection compliance at risk.
For academics conducting interviews, focus groups and qualitative research, AI transcription offers enormous advantages. Hours of recordings can be converted into text in minutes, reducing administrative workloads and allowing researchers to concentrate on analysing their findings.
However, there is a serious problem that many researchers overlook.
Where does your research data actually go when you upload a recording to an AI transcription service?
The answer may be considerably more complicated than the reassuring statements on a provider’s website.
Your Research Recordings May Contain Highly Sensitive Information
Academic interviews frequently involve confidential information about participants’ health, employment, finances, relationships, personal experiences or political opinions.
Some projects involve particularly vulnerable individuals or sensitive research subjects.
Under UK GDPR, recordings containing identifiable individuals constitute personal data. Information concerning health, ethnicity, religious beliefs and certain other matters may also qualify as special category data, requiring additional protection.
Uploading this material to an external transcription service constitutes a form of data processing that must comply with the university’s policies, research ethics requirements and applicable data protection legislation.
Researchers cannot simply assume that a service is suitable because it is popular, inexpensive or widely advertised.
The Problem With Overseas AI Transcription Providers
Many widely available AI transcription services are operated by American technology companies or use international cloud infrastructure.
This creates several questions.
Where are uploaded recordings stored? Where does the actual transcription processing take place? Can employees or subcontractors outside the UK access the recordings? Are files transferred between international data centres?
A provider might advertise European hosting while using overseas infrastructure for particular processing operations.
International transfers are not automatically unlawful. However, they must satisfy UK GDPR requirements, including appropriate transfer arrangements where applicable.
Universities may also have internal restrictions that go further than the minimum legal requirements.
For academics handling sensitive research material, establishing the location of processing and the legal basis for any international transfers is essential.
The safest approach is to establish where every stage of processing takes place before uploading a recording.
Could Your Research Be Used to Train AI Models?
Another concern is whether recordings uploaded to transcription services are subsequently used to develop or train AI systems.
Some AI providers reserve rights to use customer information for service improvement or model development, depending on their contractual terms and account settings.
Others expressly exclude customer data from training.
Researchers therefore need to examine the actual terms governing their chosen service rather than assume that all AI providers follow the same practices.
Consider an academic interviewing patients about their experiences of mental health treatment.
The recordings may contain deeply personal information provided on the understanding that it will be used only for an approved research project.
If those recordings are uploaded to a service that permits their use for AI training, the consequences could extend beyond ordinary confidentiality concerns.
The researcher may have exceeded the purposes described in the participant information sheet or the scope of the project’s ethics approval.
Once information has been incorporated into model-development processes, establishing precisely how it has been used may also become difficult.
Researchers should insist on an explicit contractual assurance that their recordings and transcripts will never be used to train AI models.
What Are UK Universities Saying About AI Transcription?
Several British universities have published specific warnings about external transcription services.
University College London
University College London warns that automated transcription services may not disclose where recordings are processed, how long they are retained or whether they are used for other purposes.
Its data protection guidance states:
“Without proper contractual arrangements in place, these aspects are not controlled and present a data protection risk to UCL.”
The university recommends considering approved transcription facilities and appropriate data management arrangements.
Source: UCL – Data Protection Frequently Asked Questions.
Loughborough University
Loughborough University published guidance in February 2026 specifically addressing the security risks associated with AI transcription tools.
It identifies international transfers, potential AI training, loss of control over recordings and inadequate consent as important concerns.
The university advises staff:
“Avoid uploading recordings with personal, sensitive or confidential information to unvetted platforms.”
It recommends using approved university facilities and obtaining appropriate information governance advice before adopting external tools.
Source: Loughborough University – AI Transcription Tools: A Time Saver or Security Risk?.
University of Bath
The University of Bath’s research data guidance is particularly relevant.
It warns that automated transcription services are often based in the United States and may present difficulties with UK data protection requirements.
Its guidance states that researchers using external transcription providers must establish appropriate data-sharing arrangements before transferring personal data.
The university also emphasises checking the provider’s location, security arrangements and contractual terms.
Source: University of Bath – Transcription Guidance.
University of Salford
The University of Salford’s ICT Acceptable Use Policy contains a particularly clear restriction.
It instructs users:
“Not enter university, staff or student data into Artificial Intelligence systems not specifically approved by the University (Digital IT).”
The policy also requires a Data Privacy Impact Assessment and prior approval for AI systems involving personally identifiable data.
This illustrates why even a secure transcription service may require university approval before researchers can use it.
Source: University of Salford – ICT Acceptable Use Policy.
A UK-Based Alternative Designed for Academic Research
These concerns have helped shape the development of Unitranscribe, a specialist AI transcription platform being developed for academic researchers.
Owned by TP Transcription Limited, an established British transcription company, Unitranscribe is being built in conjunction with an academic team at the University of Salford.
The objective is to combine the speed of AI transcription with the confidentiality and data protection standards expected by British universities.
Rather than competing primarily on price, the service is designed around four important assurances.
1. All Processing Takes Place in the UK
Unitranscribe processes recordings within the United Kingdom.
Neither uploaded audio nor completed transcripts are transferred overseas.
This provides researchers with a straightforward answer to one of the most important questions raised during university information governance assessments.
UK processing does not automatically guarantee GDPR compliance, but it removes the additional complications associated with international transfers of recordings and transcripts.
2. Recordings Are Never Used to Train AI Models
Unitranscribe does not use uploaded recordings to train AI models.
This is a fundamental commitment.
Researchers can use AI transcription without their participants’ conversations being repurposed for model training.
For academics working with confidential interviews, commercially sensitive material or vulnerable research participants, this distinction is particularly important.
3. Established Information Security Arrangements
Unitranscribe is owned by TP Transcription Limited and operates under its established information security arrangements.
TP Transcription Limited holds ISO 27001 certification and Cyber Essentials Plus certification.
These provide an established security framework supporting the development and operation of the new service.
Importantly, this does not mean that individual research projects are automatically approved for processing through Unitranscribe. Universities must still undertake any necessary supplier assessments and approval procedures.
4. Developed With Academic Input
Unitranscribe is being developed in conjunction with an academic team at the University of Salford.
This collaboration is intended to ensure that the platform reflects the practical requirements of academic research rather than simply adapting a general-purpose business transcription product.
Research interviews, focus groups, ethics requirements, confidentiality and secure handling of participant information are central considerations.
The result is a service being designed specifically for the academic market.
Researchers Should Ask Five Questions Before Uploading Anything
Before using an AI transcription platform, researchers should establish:
- Where will the recordings and transcripts be stored and processed?
- Will any data be transferred overseas or accessed internationally?
- Will recordings or transcripts be used to train AI models?
- What security certifications, contractual protections and deletion arrangements apply?
- Has the service been approved under the university’s information governance and research ethics procedures?
If a provider cannot answer these questions clearly, researchers should think carefully before uploading confidential material.
AI Transcription Should Save Time, Not Create Data Protection Problems
AI transcription has enormous potential to improve academic research.
However, speed and convenience should never come at the expense of confidentiality, research ethics or data protection.
Researchers need to understand precisely what happens to their recordings, particularly when using external AI services.
Unitranscribe is being developed to offer UK academics a secure, UK-based alternative, with transparent data processing arrangements, no AI training on uploaded recordings and the backing of an established ISO 27001-certified transcription business.
The platform is currently in development.
Academics, research departments and universities interested in its launch can visit Unitranscribe.co.uk to register their interest, receive updates and enquire about potential pilot opportunities.
Your research data deserves more than a cheap transcription service. It deserves a provider that understands why confidentiality matters.
Jonathan Fagan is MD of Ten-Percent.co.uk Limited and TP Recruitment Limited, a set of websites involved in a range of recruitment work. Jonathan has been running small businesses for over 20 years and has a number of interests and investments in companies as diverse as transcription, legal recruitment and a bit of marketing. He is an author of a number of guides and books, together with a children’s novel. In his spare time he enjoys playing golf, cricket, coaching girls’ football, operating a parent taxi, lots of running and paddleboarding on Bala Lake and the River Dee. He is a strong believer in a good work-life balance and regular blogs on making money vs enjoying life. Jonathan’s website is https://www.jonathanfagan.co.uk