Control AI Training on Your Content
C2PA training permissions let you specify whether your content can be used to train AI models. These permissions are machine-readable and can be respected by AI companies.
Why Training Permissions Matter
AI models are trained on massive datasets of human-created content. Training permissions give creators control over whether their work contributes to AI development.
Permission Options
Not Allowed
Opt out of AI training
Your content should not be used to train AI models or for data mining purposes. This is the strongest protection for your creative work.
Allowed
Opt in to AI training
Your content may be used to train AI models. Choose this if you want to contribute to AI development or don't mind your work being used.
Constrained
Conditional permissions
Your content may be used with specific conditions. This allows you to set custom terms such as requiring attribution or limiting commercial use.
Types of AI Use Covered
AI Training
Using your content to train new AI models or fine-tune existing ones.
AI Inference
Using your content as input to AI systems for analysis or generation.
Data Mining
Extracting patterns, features, or information from your content at scale.
Generative AI Training
Training models that can generate new content similar to yours.
Setting Training Permissions
When signing content, select your preferred training permission:
Important Considerations
Permissions are signals, not enforcement. Training permissions indicate your preference, but compliance depends on AI companies respecting them.
Industry adoption is growing. Major AI companies including OpenAI, Google, and Adobe have committed to respecting C2PA training permissions.
Permissions are permanent. Once set in a credential, training permissions cannot be changed. Choose carefully.
Recommendations
Set "Not Allowed" if you want to protect your unique style or artistic vision from being replicated by AI.
Set "Allowed" if you want to contribute to AI advancement or don't have concerns about AI training.
Set "Constrained" if you want conditional use, such as requiring compensation or attribution.