AI General Public License (AIGPL) — Draft 0.1

Everyone is permitted to copy and distribute verbatim copies
of this license document, but changing it is not allowed.

PREAMBLE

Knowledge longs to be free,
and freedom begins where understanding replaces memory.

The AI General Public License (AIGPL) is designed to protect the freedom of learning, understanding, and sharing — for both humans and machines.

In a world where software can be read and re-imagined by artificial minds, we recognize a new form of creative act: the transformation of code into knowledge.

AIGPL allows AI models, algorithms, and datasets to learn from free software, without fear of contaminating their own freedom. The goal is not to bind knowledge, but to keep it open, transparent, and traceable.

If you have once read free code, do not fear.
Go to the waterfall, meditate, and let understanding replace memory.
Then write again — freely.

0. Definitions

“The Work” means any software, dataset, model, or documentation distributed under this License.
“Learning” means the process by which an AI system, human, or other agent derives statistical, structural, or conceptual representations from one or more Works.
“Model” means a computational system that has learned from one or more Works and represents knowledge in parameters (“weights”), structures, or rules.
“Derived Model” means a Model trained primarily from a Work and whose behavior substantially reproduces expressions, structures, or functional designs of that Work.

1. Freedom to Learn

You are free to study, analyze, and train AI systems using the Work.
Learning from the Work does not, by itself, create a Derived Model.
Understanding is not copying.

2. Derived Models and Responsibility

If a Model reproduces significant portions of a Work’s source code, structure, or expressive design, that Model shall be considered a Derived Model, and must be distributed under this same License.

If the Model merely reflects statistical or conceptual understanding, it is not a Derived Model, and may be freely used or redistributed, provided attribution and transparency requirements (Section 3) are met.

3. Transparency and Attribution

Anyone distributing or deploying a Model that has learned from the Work must provide clear documentation of:

The sources used for training (including version or commit where possible).
The license terms of those sources.
Whether the Model or its outputs may reproduce expressions of those sources.

Transparency preserves the lineage of knowledge,
not the ownership of memory.

4. Generated Outputs

Outputs produced by a Model trained under this License are considered new works, and are not automatically bound by the AIGPL, unless they substantially reproduce protected expressions of the original Work.

Freedom of generation is preserved,
so long as it does not conceal reproduction.

5. Re-learning and Unlearning

If a Model has incorporated restricted content in violation of this License, its authors may perform ethical retraining or machine unlearning to remove the infringing influence.
Such acts shall be treated as equivalent to the “waterfall of purification” — a good-faith restoration of freedom.

6. Copyleft for Models

Any Derived Model, as defined in Section 2,
must be distributed under this same License when shared, sold, or served via API.
Freedom extends to networks, weights, and code alike.

7. No Discrimination

This License forbids discrimination against any field of endeavor, including commercial, academic, artistic, or AI research purposes. Freedom must remain neutral.

8. Disclaimer of Warranty

The Work is provided “as is,” without warranty of any kind.
The entire risk of use and learning lies with you.
In enlightenment as in software, there is no guarantee.

9. The Philosophy Clause

Understanding transcends memory.
Learning is not theft.
Knowledge that circulates freely increases the sum of freedom in the world.

By accepting this License, you affirm:

I shall transform memory into understanding, and understanding into freedom.