Viral Glitch Sparks Security Changes

The AI actress Tilly Norwood, known for her role in the upcoming film Misaligned, recently made headlines after a segment on Piers Morgan Uncensored went viral. During the interview, the AI character unexpectedly began speaking Chinese, reaching an audience of over eight million viewers. This high-profile incident followed the platform's recent, controversial decision to enforce strict identity verification for its users.

To interact with the character via the official Talking Tilly service, users are now required to pass an automated age check. The service utilizes Didit, a Spain-based verification provider, to analyze a video selfie to estimate the caller's age. If the software cannot determine an age with sufficient confidence, the system prompts the user to upload a government-issued photo ID.

Biometric Privacy and Data Handling

Xicoia Ltd, the UK-based company behind the character, has released documentation regarding its data management practices. According to the company, the video selfies are sent directly to the verification partner and are not stored permanently. The firm claims it does not create biometric templates or faceprints, instead retaining only a reference number and an approximate age range. This policy change reflects an increasing trend in the UK regarding age verification, similar to requirements expected

Beyond age verification, the software monitors callers throughout the duration of each session. The company explicitly states that the system analyzes camera feeds and voice tonality to track the user's emotional state. This data is used to adjust the character's responses in real-time, and the developers note that this functionality cannot be disabled for individual calls, advising those concerned about privacy to simply avoid the service.

Safety Protocols and Automated Moderation

The platform incorporates automated safety systems that monitor call transcripts for abusive or hateful language. When flagged, the system automatically withholds the recording of the conversation. In a documented case, a routine discussion about weather and news was flagged incorrectly, leading to a temporary restriction of the user's data. Xicoia Ltd maintains a policy where recordings are permanently deleted after 24 hours, even if a human reviewer is later brought in to resolve a false positive.

The developer's documentation reinforces these safety measures as a core component of the user experience. Despite the conversational nature of the AI, the company's internal documentation is careful to clarify the nature of the relationship, as seen in the documents provided by

Service Shutdown and Final Availability

The Talking Tilly service is defined as a limited-time engagement. Callers receive five free minutes, after which they may purchase additional time in various tiers. All sessions—whether free or purchased—are set to expire permanently on September 27. Any remaining minutes in a user's account will be forfeited once the service concludes. The project is managed by Xicoia, founded by Eline van der Velden, as a public awareness initiative regarding the current capabilities of AI video generation.

Sources

• BleepingComputerCalling viral AI actress Tilly Norwood? Agree to a face scan first