AI Actress “Talking Tilly” Service Scans Callers’ Faces and Moods Without Consent
What Happened — The UK‑based Xicoia Ltd. launched “Talking Tilly,” an AI‑driven video‑call service that forces every caller to submit a selfie for an automated age check and continuously analyses the video‑feed and voice tone to infer emotional state. The provider claims no biometric template is stored, yet the selfie and voice data are processed by third‑party Didit and Google Gemini, and the service relies on a “legitimate interests” legal basis rather than explicit consent.
Why It Matters for Trust & Control Assurance
- Demonstrates a gap in privacy‑by‑design and lawful‑basis documentation – a core control objective for continuous privacy assurance.
- Highlights the need for transparent consent management and evidence‑backed data‑minimization when processing biometric‑like data.
- Shows how AI‑driven mood‑sensing can trigger false‑positive content‑blocking, underscoring the importance of AI governance and audit‑ready logging of automated decisions.
Who Is Affected – Media & entertainment firms offering consumer‑facing AI avatars, AI‑SaaS platforms that process video/audio, and any organization that uses biometric‑like analytics for user verification.
Recommended Actions
- Conduct a privacy impact assessment (PIA) focused on biometric and mood‑analysis processing.
- Document the lawful basis (consent vs. legitimate interest) and update privacy notices to give users a real opt‑out.
- Implement audit‑ready logging of AI inference outcomes and provide a manual review pathway for false‑positive blocks.
- Map these practices to the VCF control objective “Privacy and data‑protection controls” and capture evidence in your Trust Center.
Technical Notes – The age‑check uses a selfie sent to Didit (Spain) for AI‑based age estimation; mood detection runs on Google Gemini via Tavus, with live transcription and US‑based storage. No biometric template is retained, but approximate age band and reference numbers are stored. Source: BleepingComputer