Does anyone know of active work happening on such a license?

Writing the license is the easy part, the challenge is in making it legally actionable. If AI companies are allowed to get away with "nuh uh we ran it through the copyright-b-gone machine so your license doesn't count" then licenses alone are futile, it'll take lobbying to actually achieve anything.

▲

tazjin a day ago | parent [-]

Huh? Clearly writing it is not easy, as per your own comment

	▲	jsheard a day ago \| parent [-]
		My point is that you could write the most theoretically bulletproof license in the world and it would count for nothing under the precedent that AI training is fair use, and can legally ignore your license terms. That's just not a problem that can be solved with better licenses.

▲

hoppp a day ago | parent | prev | next [-]

I got an

"LLM Inference Compensation MIT License (LLM-ICMIT)"

A license that is MIT compatible but requires LLM providers to pay after inference, but only restricts online providers, not self-hosted models

	▲	HumanOstrich 3 hours ago \| parent [-]
		That's not MIT-compatible, it's the opposite. MIT-compatible would mean that code under your license could be relicensed to MIT. Similar to how the GPL is not MIT-compatible because you cannot relicense GPL code under MIT.

▲

glerk 21 hours ago | parent | prev [-]

I can ask Claude to generate you one right now. It will be just a bunch of bullshit words no matter how much work you put into writing them down (like any other such license).