Core Consent Standard
Cryptographically-verifiable consent tokens that scope how data may be used for training, inference, agent actions, and memory, with built-in attribution limits, expiry, economics, and revocation.
01Abstract
Defines a decentralized protocol for managing consent between humans and AI systems. It establishes cryptographically-verifiable permissions for data usage in training, inference, and agent actions, ensuring no single entity controls the consent layer of AI interactions.
02Key concepts
Consent tokens
A signed token carries the owner, a permission bitmask (train, infer, agent, memory), the models and purposes it applies to, expiry, and rates.
Scoped permissions
Permissions are explicit and granular, granted per model and per purpose rather than as a blanket yes.
Attribution bounds
A maximum-influence ceiling limits how much any single source may shape a model's output, making attribution tractable.
Economics & revocation
Tokens carry training and inference rates, can be revoked, and may enable unlearning where the implementation supports it.
03Core operations
- grantConsent(dataHash, permissions, duration)
- checkConsent(user, dataHash, requestedPermission)
- revokeConsent(tokenId)
LCS-001 is the foundation of the protocol. Every other standard builds on the consent token and the grant → check → revoke lifecycle it defines.
04Full specification
The complete, canonical specification for LCS-001 lives in the standards repository and is always the source of truth.
Read on GitHub