<p>This paper revisits John Danaher and Sven Nyholm’s influential proposal of a “minimally viable permissibility principle” (MVPP) for personalised “digital duplicates.” MVPP outlines five necessary ethical conditions—consent, creation of minimal positive value, transparency, harm mitigation, and non-authenticity—that must be satisfied for the creation and use of such AI-powered twins to be ethically permissible. While recognizing the value of this framework, this paper argues that ethical analysis must also engage with the legal and data architecture that enables the responsible creation and operationalization of such synthetic entities. Drawing on prior scholarship, it introduces three complementary legal principles for structuring the technological infrastructure: a human-centric approach to regulating personal data, a private-by-default data architecture for AI systems, and robust data dominion rights for individuals. Together, these principles reframe personal digital duplicates as extensions of personal identity rather than assets controlled by corporations. As such, the paper advocates a new human-centric social contract with technology for the AI age—one that combines the MVPP’s ethical safeguards with enforceable data rights and privacy-preserving technical architecture. By placing individuals at the center, such a framework seeks to better align innovation with human dignity, autonomy, and trust, ensuring that personal digital duplicates are integrated into society responsibly, on terms that best preserve individual sovereignty and control.</p>