Can I train an AI model on CDLA Permissive 2.0 material?

Short answer: Training a model: Yes. Training a model you use commercially: Yes.

Community Data License Agreement Permissive 2.0 (CDLA-Permissive-2.0): Use, modify and share the data for any purpose; when sharing the data, include the agreement. Results of computational use, such as trained models, carry no restrictions or obligations.

Training a model: Yes

License: Community Data License Agreement Permissive 2.0 (CDLA-Permissive-2.0, curated)

• Every use you described is granted by CDLA Permissive 2.0, and nothing you described is prohibited.

Duties: license_notice

Training a model you use commercially: Yes

License: Community Data License Agreement Permissive 2.0 (CDLA-Permissive-2.0, curated)

• Every use you described is granted by CDLA Permissive 2.0, and nothing you described is prohibited.

Duties: license_notice

Context

• Whether a trained model is a copy or adaptation of its training data is legally unsettled; where the license does not settle it, Rightshold marks training for human review instead of guessing.
• Written for data and machine learning: the agreement places no restrictions or obligations on results such as trained models.
• Sublicensing is not documented.

Sources

• Community Data License Agreement Permissive 2.0: official text
• CDLA-Permissive-2.0 on the SPDX License List

Check it yourself (free, no key)

{
  "tool": "check_license",
  "arguments": {
    "license": "CDLA-Permissive-2.0",
    "intended_use": {
      "model_training": true
    }
  }
}

Computed from Rightshold's license profiles; not legal advice. Updated 2026-10-19. JSON · Markdown