Research
GeraMind is building the personal context layer for AI: a sovereign, encrypted vault that lets users decide which AI systems can query which facts about them, under what conditions, and for how long. Our research programme addresses the hardest unsolved problems in that design space — from consent scoping to purpose-binding, from cryptographic graph structures to the regulatory frameworks that must evolve to protect individuals in an agent-saturated world.
Research Themes
- Consent-scoped query design. How should a personal context vault respond when an AI agent requests a user's dietary preferences to complete a restaurant booking? GeraMind research defines a formal consent model with purpose binding, temporal expiry, and minimum-disclosure principles — ensuring that agents receive exactly the context they need for a specific task and nothing more.
- Personal context graphs. Facts about a person form a graph, not a flat key-value store. Relationships between nodes (health condition → medication → pharmacy → insurance plan) create inference risks that naive data minimisation cannot address. We develop graph-theoretic privacy metrics and redaction algorithms that preserve utility while closing re-identification paths.
- Interoperability with AI memory systems. ChatGPT Memory, Claude's Projects context, and MCP memory servers all implement competing personal context primitives. GeraMind research maps these systems, identifies their trust assumptions, and proposes a migration protocol so users can move their context between providers without losing fidelity or control.
- Data sovereignty regulation. GDPR Article 22 (automated decision-making), the EU AI Act's transparency obligations, and emerging personal-data-as-property frameworks in US state law all intersect with personal context vaults. We contribute policy analysis on how GeraMind can operate lawfully across 50+ jurisdictions from day one.
First Articles — Q3 2026
The GeraMind research series launches in Q3 2026. First articles include a deep dive on consent-scoped query semantics, a comparison of personal context vault architectures (GeraMind vs Solid/Inrupt vs Apple Data Vault), and an analysis of the re-identification risk surface in personal context graphs.
To be notified when new research is published, join the GeraMind waitlist. You can also explore our current thinking in the GeraMind blog.