ScotRail's AI Train Announcer Was Built From a Real Artist's Voice Without Her Knowledge
What happened
In May 2025, Scottish railway operator ScotRail began rolling out a new AI-driven train announcer it called "Iona," a synthetic voice developed by Swedish technology company ReadSpeaker. The launch attracted positive coverage for its local character, a voice with a Scottish accent built specifically for the network. What the coverage did not include was the name of the person whose recordings may have made it possible.
Gayanne Potter, a professional voiceover artist, recognized her own voice in the announcements and came forward to say she had never agreed to any of it. Potter had recorded material for ReadSpeaker in 2021 under what she understood to be a narrow brief: the recordings would be used for accessibility tools and e-learning products. She was not told her voice would be used to train a synthetic announcer for public transport, and she did not consent to this commercial use. She and her agent say they were explicitly assured the recordings would not be sold to third parties.
ReadSpeaker disputes that account. The company maintains that its contract with Potter permits the use of synthesized voices for businesses and organizations, and that ScotRail's deployment of "Iona" falls within those terms. Both sides are working from the same document and arriving at opposite conclusions, which is itself the problem. A contract signed in 2021 for one stated purpose was applied four years later to a substantially different commercial deployment, and nothing in the UK's current legal framework clearly prevents that from happening. Unlike the United States, which has state-level right-of-publicity laws that cover voice likenesses, the UK has no specific statute protecting individuals from having their voices cloned and deployed commercially without their ongoing consent.
The impact on Potter has been direct. She describes feeling violated and devastated, and she has raised concerns that the AI version of her voice will compete with her real work in the professional market she has built over her career. An AI voice trained on her recordings and deployed at scale across a national rail network represents a form of commercial competition that no contract written in 2021 for an e-learning brief was designed to address.
The case surfaces a gap that has nothing to do with any single company or technology vendor. When a public-facing voice system is built on identifiable recordings of a real person, there is currently no requirement to document that derivation, disclose it to the person involved, or prove what the original consent actually covered. A provable record of what a system was built on, and under what terms, would shift the burden of proof away from the individual who has to discover the harm on their own, often long after it has already been deployed.
Reported impact
- Affected parties
- Not publicly disclosed
- Harm type
- Not publicly disclosed
- Scale
- Not publicly disclosed
- Financial impact
- Not publicly disclosed
- Regulatory action
- Not publicly disclosed
Classification
Relevant governance controls
Governance control mapping is not available for this record.
- No controls mapped
Not publicly disclosed
Control mapping is analytical. It does not state that any control would have prevented the incident.
Sources and evidence
This record was researched and written by the Index. The event is also catalogued in the following database, which is listed for cross-reference.