Who controls the memory of your AI system, and why does that question matter for democracy, sovereignty, and the future of digital governance? Most modern AI systems collocate computational power with long-term user data inside centralised cloud infrastructure. Such a design choice has made the centralisation of user memory one of today's most pressing AI governance challenges. This session proposes an alternative: AI systems can operate as stateless reasoning engines while user memory is governed through distributed, user-controlled architectures. We examine the regulatory implications of this unbundling under existing regulatory frameworks (including the EU Digital Markets Act) and explore the geopolitical difficulties facing middle powers seeking to coordinate standards and protect digital sovereignty. The window to shape these architectural choices is narrowing. As AI systems become embedded in public services, financial systems, and everyday life, the governance frameworks that determine who owns, accesses, and controls user memory will have consequences that extend far beyond technology policy into questions of power, accountability, and the capacity of middle powers to protect their citizens' digital sovereignty on their own terms.