[License-review] For Approval: OpenMDW License Agreement, versions 1.1 (OpenMDW-1.1)

[Stefano Maffulli]

On 8/24/26 03:20, Pamela Chestek wrote:
>
> Taking AI entirely out of the picture, you confirmed that the 
> copyright owner of an MIT-licensed package would be precluded from 
> bringing a lawsuit against the OpenMDW-1.1 licensor for that 
> licensor's failure to comply with the terms of the MIT license. So 
> what you are proposing is that a completely non-culpable party has to 
> give up a claim against what might be a deliberate, intentional, 
> unlawful act, which is not the patent scenario.
>
I don't understand how you get to this conclusion. Are you arguing that 
every single piece of content ingested to train a LLM must be explicitly 
licensed? If so, how do you square it with the US fair use principles 
and the text and data mining exception in the EU?

My impression reading the license and dealing with AI and the OSD is 
that the license is actually trying to protect open developers: the most 
transparent models either "suck" because they don't have enough openly 
licensed data to train on or are immediately punished by lawsuits 
because they disclose their training sets (and get sued, to test 
copyright law boundaries or to extract some revenue).

This dilemma is the center of the whole open source ai definition debate 
and IMO won't be solved by a license but needs to be solved at the 
policy level. IMO the evaluation of the OpenMDW license (or any new 
license that addresses the space) shouldn't be encumbered by policy 
discussions.