Key terms in this module(9)
Quick definitions for the terms used here. Open any term for the full entry.
- Article 53
EU AI Act Article 53. Governs general-purpose AI provider obligations including training-data documentation and TDM opt-out respect.
- CDR
Core Data Record. A structured record capturing key information about a creative asset, including the rights, consent, and provenance information attached to it. The CDR is the primary provenance infrastructure for AI pipeline entry.
- cip.md
The framework's declaration file format. A plain-text Markdown file at https://[domain]/.well-known/cip.md declaring rights, consent, and certification status in machine-readable form. Specified in the v3.18 consolidated cip.md Specification page; AI Generation field set added v3.24.
- DUA Act
Data (Use and Access) Act 2025 (UK). Establishes the UK's post-Brexit data governance framework including the AI training opt-out provisions. Requires either licence or explicit permission for commercial AI training; rights holders may opt out using machine-readable signals.
- EU AI Act
Regulation (EU) 2024/1689. Establishes the EU's comprehensive AI regulation framework. Article 50 covers AI-output disclosure obligations on deployers; Article 53 covers training-data documentation and TDM opt-out respect by general-purpose AI providers.
- Inference
Real-time computation of outputs from inputs using learned parameters. The operational stage where AI model meets specific input to produce a specific output.
- Source material
Third-party copyrighted material (other than trademarks) present in the operator's content but owned by parties other than the operator. Categories include quotations and excerpts (text from books, articles, speeches), audio samples (music production sampling, sound effects), film/TV clips (documentary footage, criticism content), archival material (historical footage, period photographs), stock material (commercial stock photography, library music), performances (cover songs, sampled performances, archival performance footage), quoted dialogue and characters (fan fiction, parody, transformative works), datasets and structured material (licensed databases, academic datasets), and software and code (third-party libraries, open-source incorporation). Declared via CIP-Source-Material (v3.34). Distinct from operator-owned material handled by the standard rights bundle, and from third-party trademarks handled by CIP-Third-Party-Marks (v3.33).
- TDM opt-out
Text and Data Mining opt-out. A declaration by a rights holder that their content may not be used for AI training. Under UK Data Use and Access Act 2025 and EU AI Act Article 53, machine-readable opt-out mechanisms carry distinct legal weight.
- Temporal consent
Time-limited consent that automatically lapses after a declared period. After expiry, consent reverts to the framework's default-restrictive position; the rights holder must be approached again. Codified in CIP-Consent-Expires field.
3.1 The consent matrix
Different AI uses require different consents. The CIP consent matrix defines five distinct consent types:
AI training ingestion — CIP-Training-Consent: Granted authorises copying content into a training corpus.
Fine-tuning — Fine-tuning consent is separate from training consent and authorises using content for targeted model adaptation.
Inference / retrieval — CIP-Prompt-Use: Permitted authorises AI reading and referencing content in responses.
Style transfer — CIP-Style-Transfer: Permitted authorises AI learning and applying creative style.
Synthesis — CIP-Synthesis: Permitted authorises AI generating content using the work as source material.
Most creators seeking full protection should set CIP-Training-Ingestion: Prohibited and CIP-Fine-Tuning: Prohibited. Setting CIP-Prompt-Use: Permitted is generally reasonable — it allows AI assistants to reference your work in responses without training on it.
3.2 Temporal consent
Temporal consent is a time-limited authorisation. After the period expires, the consent lapses automatically. The AI company must either seek renewed consent or cease using the content.
3.3 Scope-limited consent
Scope-limited consent authorises a specific type or quantity of use rather than a general permission — non-commercial research only, one jurisdiction only, one named AI company only, a specific model version only. Any use outside the defined scope creates subsisting rights failure liability.
3.4 The TDM opt-out
The four-field combination required for a complete, legally supported TDM opt-out:
- CIP-TDM-Opt-Out: true
- CIP-TDM-Opt-Out-Scope: All-Content
- CIP-TDM-Jurisdiction: UK, EU, US
- CIP-Training-Ingestion: Prohibited
Under UK DUA Act 2025 and EU AI Act Article 53, a machine-readable opt-out creates a legally recognised basis to object to commercial training use. A TDM opt-out can be withdrawn for future use — update your cip.md and CDR to Not-Granted — but withdrawal does not retroactively authorise what has already been done.
Summary
Key Takeaways
- Consent must be machine-readable, specific, informed, and revocable
- Training and fine-tuning require separate consents — one does not cover the other
- Temporal consent automatically expires and requires renewal
- The four-field TDM opt-out combination is the minimum for a legally supported opt-out
- Withdrawal prevents future use but does not undo past training
Self-check
Check Your Understanding
- Which two AI use types require separate, distinct consents under the CIP framework?
- What happens when temporal consent expires?
- How many fields are required for a complete, legally supported TDM opt-out in cip.md?
- A creator declares a TDM opt-out today. An AI company trained on their work last year. Does the opt-out create liability for the past training?
- What is scope-limited consent?