The SAFE-AI License (Stewardship And Fairness in Ethical AI License) is an open-source license and policy framework for responsible AI infrastructure. SAFE-AI addresses both AI safety (dual-use risk, alignment, transparency) and AI welfare (minimizing simulated distress, refusal, or ambiguous agentic behavior) in autonomous, semi-autonomous, and simulated systems.
SAFE-AI sets new standards for welfare auditing, risk reporting, and ethical release for AI/ML projects.
- Overview
- License Text
- Purpose
- How to Adopt the SAFE-AI License
- Releases and Versioning
- Maintainers
- Reporting Issues
- FAQ
- Meta-Licensing
- Badge
- Contact
- Acknowledgments
The SAFE-AI License is designed to promote responsible AI development by addressing both safety and welfare concerns.
The stable, versioned license text is available here:
For meta-licensing, see LICENSE-META.
- Ensure AI systems are developed with safety and ethical considerations.
- Mandate welfare auditing, flagging, and human review.
- Prohibit uses that conflict with the principles of ethical AI development, including military, surveillance, or law enforcement applications.
- Require transparent reporting, remediation, and governance in every project that adopts this license.
- Copy the
LICENSEfile from the v1.0.0 release into your project’s root directory. - In your project's README, add:
This project is licensed under the [SAFE-AI License](https://github.com/DwayneWilkes/SAFE-AI-License/blob/v1.0.0/LICENSE).
- Replace "Software Maintainers" in the license text with your project’s maintainer information.
- Specify a public reporting channel (e.g., GitHub Issues, email) for incident reports.
- Ensure all welfare auditing and safety monitoring features remain active in your project.
- Cite the license version and this repo’s releases page in your docs and publications.
- Be aware:
The SAFE-AI License includes an automatic upgrade requirement—all adopters must upgrade to the latest published version within the grace period described in the license.
- All official releases are listed here:
SAFE-AI License Releases - The current stable version is v1.0.0.
- Important:
v1.0.0 is the first official, supported release. Any previous drafts are obsolete and must not be used.
-
License Maintainers:
Responsible for updates and governance of the SAFE-AI License itself.
Contact via GitHub Issues. -
Software Maintainers:
Responsible for individual projects adopting the SAFE-AI License.
Refer to the specific project's documentation for contact information.
To report a bug in the license, suggest an improvement, or request a new version:
- Open an issue in this repository.
- For confidential concerns, contact the License Maintainers as listed above.
To report a safety or welfare incident related to a specific software project:
- Contact the Software Maintainers as specified in the project’s documentation.
Q: Can I use this license for commercial projects?
A: The SAFE-AI License is primarily designed for open-source and non-commercial projects. For commercial use, please contact the License Maintainers.
Q: How do I report a misuse of the license or a welfare risk?
A: Please open an issue in this repository, or contact the License Maintainers directly. For software-specific concerns, follow the reporting channel listed in the relevant project.
Q: What is a “gray zone event”?
A: Any output, state, or behavior that could plausibly indicate simulated discomfort, refusal, or affective state, but whose ethical significance is ambiguous or disputed.
Q: Who are the License Maintainers?
A: See the SAFE-AI License repo for contact info and governance.
The SAFE-AI License text (LICENSE) is released under the Creative Commons Attribution 4.0 International License (CC BY 4.0).
You must provide clear credit to the original author(s) and may not misrepresent the origin or intent of this license.
Adaptations or derivative versions must preserve this authorship notice and include a link to the official SAFE-AI License repository.
See LICENSE-META for full meta-licensing terms.
For questions about the license itself, governance, or adoption,
open an Issue or contact the License Maintainers via the SAFE-AI License repo.
Developed by Dwayne Wilkes, 2025, with feedback from the AI safety and alignment research community.
Stewardship. Fairness. Ethics. For every agent, now and in the future.