CC license is fundamentally incompatible with generative AI.

[greg-kennedy]

Description

It seems fundamentally impossible to use generative AI trained on CC licensed content in a manner compatible with the original licensing. Consider:

1. A CC license can be considered a disclaimer of some, but not all, rights provided by copyright laws. For example, it allows sharing, but with conditions (i.e. "name must stay attached", "do not alter this", etc). These retained rights are protected by the copyright infrastructure, and dependent on its protections to function.
2. AI model output is not copyrightable (at least in the U.S. and other major jurisdictions).
3. Therefore, it is impossible to enforce the derivative terms provided by a CC license when creating AI output. The AI process effectively "launders" the restrictions, producing output that legally cannot be compelled to have the same protections guaranteed by the license of step 1.

Placing a CC license on a work should not be removable by loopholes - technical, legal, or otherwise. "Signalling" to an AI scraper that the content might be OK to Scrape does not render the license moot. It does nothing to resolve the basic issue at stake here.

Expectation

My expectation is that the foundation which devised, promoted, and enforces the CC license (creative commons dot org) should not willfully participate in trying to find means to circumvent the protections its very licenses offer, but instead should stick to its mission statement and goals.

Resolution

See issue #10