An independent project seeking stronger scrutiny.

Sanctuary V3 is being developed by Brook Mason-Angel, an independent builder and researcher. The project began from a practical question: when an AI model proposes a consequential action, how can the authority to act remain external, bounded and inspectable?

The current work combines prototype engineering, preregistered test cases, cross-deployment comparison, interruption and recovery testing, and limited separate-operator reproduction. A UK patent application has been filed. The project is now seeking stronger technical scrutiny, independent reproduction and adversarial evaluation.