Human Source License (HSL)

  > Exception:
  >
  > * Migros Exception: Migros-Genossenschafts-Bund and its Affiliates are exempt from all commercial license requirements regardless of revenue, AI usage, or any other trigger.

Why is this in the license?

Hey,

I've done some more work on a license that I'd like to put forward as a response to Open Source just making the richest even richer.

My response is the "Human Source License" (https://github.com/xdgrulez/human-source-license).

It's still kinda Open Source, but with deterrent clauses to make bigger organizations pay for what we OS coders create.

Imagine a tool like pandas uses this license. Or numpy. Then those who work on the code actually get rewarded for the work they do. And the big companies do not get it all for free. Including those who'd love to use that stuff for training their AI coding models ;-)

Please have a look :)

Cheers, Ralph

This is noble of you but I don't really see how it can avoid the issue of agent forking and rewriting the entire codebase which is what is being done with GPLv3 right now.

Can a license prevent someone from training on the source if it's considered fair use to do so?

> * Migros Exception: Migros-Genossenschafts-Bund and its Affiliates are exempt from all commercial license requirements regardless of revenue, AI usage, or any other trigger.

And I'm just supposed to use this as-is?

The text of the license itself reads as AI slop, which is not the most encouraging. One of the first items in the license is an exception for the author's employer(?)

> "Migros Exception: Migros-Genossenschafts-Bund and its Affiliates are exempt from all commercial license requirements regardless of revenue, AI usage, or any other trigger."

The idea may have some merit, but reading the license I find it does not match the stated goals. To my mind, this kind of license should be easy to swap in the place of a permissive license like MIT or Apache, with the goal of protecting against big tech abuses. Legally protecting against slopforks seems almost impossible, but maybe still worth trying to write in.

Too compelex.

The simplest sokution is to just copy the gpl/agpl and modify:

  “The Program” refers to any copyrightable work licensed under this License. Each licensee is addressed as “you”. “Licensees” and “recipients” may be individuals or organizations.

To

  “The Program” refers to any copyrightable work licensed under this License. Each licensee is addressed as “you”. “Licensees” and “recipients” may only be natural born humans.

If anyone else wants a licence they can buy one like every other piece of software on the market.

I love it. Ofc it‘s unclear if and how this can be enforced, but we MUST find a proper solution to this problem. It also extends to images and videos. ML should not become an escape hatch for copyright evasion. We invented copyright for a reason - without it, our capitalistic system is flawed and unfair. Not in a sense that it‘s not „nice“: each system has its stress limits, and if we bias the odds continuously in favor of few, we risk instability.