Shufti has received U.S. Patent No. 12,749,342 B1 for its age verification technology covering an AI-based system that uses on-device facial analysis, split inference (SI), and encrypted One-Time Age Tokens (OTATs). Split-inference technology uses distributed neural computation to split AI processing between the user’s device and a secure verification server, optimizing efficiency and privacy. In split-inference models, edge devices process the initial layers before transmitting intermediate representations to a server. For age assurance, Shufti says, when a user begins an age check, its AI models analyze facial characteristics to estimate age and assess liveness locally, as a defense against spoofs and other presentation attacks. The device then generates an encrypted One-Time Age Token containing intermediate AI outputs rather than the facial image. The server receives and validates the token, finishes AI processing, and determines whether the user meets the required age threshold. Once the server completes verification, the business receives an age eligibility result without access to the user’s facial image or full identity details. The one-time token can be restricted to specific merchants and marked as used to prevent reuse.

The SI process makes it harder to reverse-engineer the age verification.