Re: [RFCv3] Counter-Proposal -- Interpretation of DFSG on Artificial Intelligence (AI) Models

On Saturday, May 10, 2025 11:03:31 AM Mountain Standard Time Aigars Mahinovs 
wrote:
> If your entire proposal is based on this assumption about how
> copyright and copyright law works, I would expect something more
> substantial, like court decisions supporting this radical new
> interpretation. And overturning things like Article 4 of EU Directive
> 2019/790 granting near complete copyright exception to text and data
> mining. This was explicitly referred to in EU AI Directive in context
> of the use of training data. And overturning of a *ton* of already
> decided "fair use" cases in USA.

I know this isn’t a ton of already decided fair use cases, but I saw the 
following link this morning.

https://www.yahoo.com/news/us-copyright-office-thoughts-ai-235936541.html

Two quotes from the US Copyright Office.

“Although it is not possible to prejudge the result in any particular case, 
precedent supports the following general observations,” the office said. 
“Various uses of copyrighted works in AI training are likely to be 
transformative. The extent to which they are fair, however, will depend on 
what works were used, from what source, for what purpose, and with what 
controls on the outputs — all of which can affect the market.”

“When a model is deployed for purposes such as analysis or research — the 
types of uses that are critical to international competitiveness — the outputs 
are unlikely to substitute for expressive works used in training,” the office 
said. “But making commercial use of vast troves of copyrighted works to 
produce expressive content that competes with them in existing markets, 
especially where this is accomplished through illegal access, goes beyond 
established fair use boundaries.”


As has already been said a number of times in these threads, it isn’t a 
settled issue whether copyrighted words can be used to train an AI model 
without the permission of the copyright holder.  The answer to that question 
probably depends on if such training ends up being considered fair use.  The 
above quotes indicate that, according to the US Copyright Office, whether it 
is fair use could depend on how the AI is used *after it is trained*.

-- 
Soren Stoutner
[email protected]

-----BEGIN PGP SIGNATURE-----
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=DBsp
-----END PGP SIGNATURE-----