Values.md – file format for personal ethical alignment (values.md)

Values.md is a proposed simple file format for encoding a person’s ethical preferences so AI agents—whether in the browser, on mobile, or embodied in robots—can make decisions that align with the user’s values rather than opaque defaults set by platform providers. Modeled on the idea of CLAUDE.md for coding preferences, VALUES.md would let users declare principles, constraints, priorities and context-sensitive rules that agents can parse and apply when acting on their behalf (hiring decisions, travel bookings, recommendations, etc.). Technically, VALUES.md raises opportunities and challenges: it enables user-level specification, machine-readable policy snippets, and interoperability across agent ecosystems (agents can ingest, weigh, and resolve value conflicts), but also requires formal semantics, provenance, signing, and enforcement mechanisms to prevent misuse or fragmentation. For the AI/ML community this is significant because it shifts alignment work from monolithic model updates to modular, auditable value specifications—opening paths for standardized formats, verification tools, and policy-aware agent architectures —while also demanding research on conflict resolution, context modeling, and security to ensure user-specified values are interpreted safely and reliably.