If you want a single, vivid example of how not to use AI on the phone, the UK regulator has handed you one. And the enforcement hasn't stopped.
What happened
The ICO fined two firms a combined £550,000 for unlawful automated marketing calls. Green Spark Energy (Durham) was fined £250,000 after making around 9.5 million calls; Home Improvement Marketing (Pembrokeshire) was fined £300,000 and ordered to stop.
The detail that matters for anyone building AI calling: the firms used "avatar" software. Recipients believed they were talking to a UK person — "Jo, Helen or Ian" — but those were scripted lines recorded by voice actors, played by call-centre agents based overseas. The technology existed to make an automated, offshore operation sound like a friendly local human. Nearly 500 people complained to the ICO or the Telephone Preference Service, many of them elderly or vulnerable.
Why it broke the rules
Under PECR, automated marketing calls require prior, informed consent — which these firms didn't have from the people they dialled. The ICO's Head of Investigations, Andy Curry, made the point that endures no matter how good the tech gets:
Advances in technology may make it harder for people to spot automated calls — but the rules don't change.
And this isn't a one-off. In July 2026 the ICO fined two more home-improvement firms £370,000 for nuisance calls targeting vulnerable people. The regulator is actively watching this space.
The lesson for legitimate AI calling
It would be easy to read this as "AI calling is risky." The opposite is true: what got punished was deception, not automation. Every failure here has a simple, lawful inverse — and it's exactly how a responsible AI caller should operate:
- Say what you are. Open with "I'm an AI assistant." (This is also where EU and UK disclosure rules are heading.)
- Have a lawful basis. Consent for automated marketing; and remember a genuine research or satisfaction survey isn't marketing at all.
- Be reachable. Present a valid UK number, screen against the TPS for live marketing, and don't hide where the call comes from.
The firms in this case spent real money engineering a lie. Doing it honestly is cheaper, and it's the only version that survives contact with the regulator.
🎧 We unpack this — and the 2026 funding wave — in a short audio briefing, above.
Related: Do you have to tell people it's an AI? · Compliance & trust · The voice-AI funding wave