The AI actress Tilly Norwood went viral this week after glitching mid-interview on Piers Morgan Uncensored and briefly speaking Chinese. Less noticed is what happens before you even get to talk to her: a mandatory face scan and continuous mood analysis, added to the “Talking Tilly” service’s terms this month.
An age check you cannot skip
Before a call connects, callers must pass an automated age estimation. A video selfie is analyzed by Didit, a Spain-based identity verification provider, with a government ID upload as fallback if the estimate is unclear. Xicoia Ltd, the UK company behind Tilly, says the selfie goes directly from the caller’s device to Didit, that no biometric template is created, and that neither the selfie nor any ID image is retained, only an approximate age band and a reference number.
The check applies globally and cannot be bypassed. It arrived alongside a workplace-use ban and new automated safety systems, all added to the terms this month.
Always-on emotion tracking
Every call also runs live analysis of the caller’s camera feed and voice tone to infer emotional state, so the character can respond to match the caller’s mood. Xicoia’s privacy policy states this monitoring “cannot be switched off for an individual call.” Both the age check and the mood-sensing rely on legitimate interests rather than consent as their legal basis, a change the company’s own version history dates to September.
Calls are recorded, transcribed, and processed live by US-based providers, with responses generated using Google’s Gemini model via the Tavus conversational video platform. An automated classifier screens each transcript for abusive language and withholds the recording if flagged; one reporter’s benign call was wrongly flagged for “hateful or abusive language” before a human reviewer released it as a false positive. Recordings are deleted after 24 hours regardless of outcome.
A short-lived service with long retention
The first five minutes are free, after which pricing ranges from £0.99 for five minutes to £22 for thirty, capped at 35 purchased minutes per person. All minutes expire when Talking Tilly shuts down permanently on September 27. Transcripts are kept for up to eight weeks and may be reviewed by staff and third-party partners, and the character retains memory of prior conversations unless deletion is requested.
Regulatory context
The biometric age gate mirrors the UK’s broader push toward mandatory age verification. Adult sites serving UK visitors have required ID or facial age estimation since July 2025 under the Online Safety Act, and an under-16 social media ban takes effect from spring 2027, with the government specifically flagging AI companion chatbots for 18+ enforcement, despite Xicoia’s own documentation stating “Tilly is not a companion.”