Appendix C — Consent Framework Chart
Comparing consent architectures for AI training on creator works
| Consent / Control Mechanism | Default State | Timing of Choice | Scope of Control | Individual Creator Control | Prospective Effect | Binds Partners & Transfers | Counts as Consent (Under This Framework) |
|---|---|---|---|---|---|---|---|
| Explicit Opt-In (Checkbox at Upload or Onboarding) | No training unless creator acts | Before training | All generative AI uses | Yes | Yes | Yes (if stated) | |
| Granular Opt-In (Per-Work or Per-Use) | No training unless creator acts | Before training | Specific works or uses | Yes | Yes | Yes (if binding and comprehensive) | |
| Default Opt-Out (Account-Level Toggle) | Training occurs by default | After training has begun | Usually limited or undefined | Partial | Sometimes | Rarely | |
| Limited Opt-Out (Feature- or Purpose-Specific) | Training occurs by default | After training has begun | Narrow (e.g., “AI features”) | Partial | Sometimes | Rarely | |
| Dataset-Level Exclusion (Non-User-Facing) | Training occurs by default | Unclear or indirect | Unclear | No (non-user-facing) | Unclear | Unclear | |
| Publisher- or Platform-Level Control Only | Training occurs by default | Outside individual control | Site-wide or collective | No | Sometimes | Sometimes | |
| Disclosure Without Choice | Training occurs by default | No choice offered | All uses | No | No | No | |
| No Disclosure, No Control | Training occurs by default | No choice offered | All uses | No | No | No |
How to Interpret This Chart
This chart illustrates how different consent and control mechanisms function in practice. It does not evaluate platforms, assign grades,
or determine legal sufficiency. “Counts as consent” reflects this Guide’s methodological definition, not a legal determination
under any statute or doctrine.
Under this framework, consent requires an affirmative choice made before AI training occurs, with training disabled by default unless
the creator actively permits it. Mechanisms that assume permission by default—such as opt-outs, toggles, or disclosures without choice—may offer limited
control, but they do not constitute consent.
Platforms may implement multiple mechanisms simultaneously. Platform grades reflect the overall consent architecture, not the presence
of any single control. This chart is therefore a conceptual reference explaining the logic behind the grading framework, not a standalone scorecard.
