Data Deed

Leverages established frameworks like Creative Commons to define permissions in a universally understood vocabulary. A deed explicitly states if data can be used for commercial AI training, non-commercial research, or derivative works.

• Uses a three-layer design: Legal Code, Human-Readable Deed, and Machine-Readable Metadata.
• Eliminates ambiguity for crawlers parsing robots.txt or TDM Reservation Protocols.

Goes beyond simple 'allow' or 'deny' to specify exact computational use cases. A deed can grant a license for Text and Data Mining (TDM) while explicitly prohibiting synthetic data generation or model weight fine-tuning.

• Distinguishes between input ingestion and output generation rights.
• Prevents unauthorized use in Retrieval-Augmented Generation (RAG) pipelines.

Embeds structured metadata (JSON-LD or RDFa) that AI crawlers and ingestion pipelines can parse without human intervention. A bot encountering a deed can automatically sort content into permissioned corpora or blocked lists.

• Integrates with Consent Management Platforms (CMPs) to syndicate preferences.
• Creates an auditable provenance chain linking data to its license.

Utilizes Content Credentials (C2PA) to cryptographically bind the deed to the specific digital asset. This tamper-evident seal ensures the license cannot be stripped or altered during scraping.

• Verifies the data lineage from origin to ingestion point.
• Provides a forensic basis for AI copyright compliance and takedown requests.

Defines the legal jurisdiction and data sovereignty requirements governing the data. A deed can restrict processing to specific geographic regions to comply with GDPR or other localized regulations.

• Specifies cross-border transfer restrictions for training infrastructure.
• Enforces purpose limitation by legally binding the licensee to the stated terms.

Establishes a protocol for revoking previously granted AI training rights. A deed can include a callback URI or signal that triggers model unlearning requests when a license expires or consent is withdrawn.

• Supports the Right to Object and Right to Erasure in automated systems.
• Facilitates compliance with storage limitation principles by setting data retention timers.

A machine-readable protocol enabling content owners to declare that their copyrighted works are reserved for Text and Data Mining, overriding general crawling permissions. Unlike a data deed's affirmative grant of rights, a TDM opt-out serves as a negative declaration—a reservation of rights communicated to automated agents through technical signals rather than legal instruments.

A standardized, auditable digital record provided to a data subject detailing the specifics of a consent transaction, including whether permission was granted for AI model training. When paired with a data deed, consent receipts serve as the evidentiary backbone—providing timestamped proof that specific usage rights were granted under defined terms.

A tamper-evident metadata structure standardized by the C2PA that attaches cryptographically signed provenance information to digital content. Content credentials complement data deeds by providing cryptographic proof of origin—ensuring that the asset referenced in a deed is authentic and its chain of custody is verifiable before training ingestion occurs.

A curated collection of training data composed exclusively of content with verified licensing agreements or explicit creator consent. Data deeds serve as the atomic legal instruments that populate permissioned corpora—each deed representing a granular, machine-readable grant that collectively forms a legally defensible training dataset.

A legally binding contract between a data controller and processor stipulating the scope, purpose, and security measures for data handling, including explicit prohibitions on secondary AI training. While a DPA governs the processor relationship, a data deed governs the data asset itself—specifying usage rights that persist regardless of which processor handles the data.

A legal constraint requiring that data collected for one explicit purpose cannot be repurposed for incompatible secondary uses without new consent. Data deeds operationalize this principle by providing a machine-readable purpose specification—explicitly enumerating permitted AI training contexts and prohibiting all others, making purpose limitation enforceable at ingestion time.