←back to thread

Meta says it won't sign Europe AI agreement

(www.cnbc.com)
324 points rntn | 3 comments | 18 Jul 25 17:56 UTC | HN request time: 1.187s | source
Show context
vanderZwan ◴[18 Jul 25 18:22 UTC] No.44608123[source]▶
I admit that I am biased enough to immediately expect the AI agreement to be exactly what we need right now if this is how Meta reacts to it. Which I know is stupid because I genuinely have no idea what is in it.
replies(5): >>44608204 #>>44608531 #>>44610646 #>>44610843 #>>44611758 #
mhitza ◴[18 Jul 25 18:30 UTC] No.44608204[source]▶
There seem to be 3 chapters of this "AI Code of Practice" https://digital-strategy.ec.europa.eu/en/policies/contents-c... and it's drafting history https://digital-strategy.ec.europa.eu/en/policies/ai-code-pr...

I did not read it yet, only familiar with the previous AI Act https://artificialintelligenceact.eu/ .

If I'd were to guess Meta is going to have a problem with chapter 2 of "AI Code of Practice" because it deals with copyright law, and probably conflicts with their (and others approach) of ripping text out of copyrighted material (is it clear yet if it can be called fair use?)

replies(1): >>44608337 #
jahewson ◴[18 Jul 25 18:43 UTC] No.44608337[source]▶
> is it clear yet if it can be called fair use?

Yes.

https://www.publishersweekly.com/pw/by-topic/digital/copyrig...

Though the EU has its own courts and laws.

replies(2): >>44608417 #>>44611760 #
dmbche ◴[18 Jul 25 18:52 UTC] No.44608417[source]▶
District judge pretrial ruling on June 25th, I'd be surprised this doesn't get challenged soon in higher courts.

And acquiring the copyrighted materials is still illegal - this is not a blanket protection for all AI training on copyrighted materials

replies(3): >>44610548 #>>44613220 #>>44616692 #
1. rpdillon ◴[19 Jul 25 16:08 UTC] No.44616692[source]▶
Why is acquiring copyrighted materials illegal?

You can just buy books in bulk under the first sale doctrine and scan them.

replies(1): >>44617526 #
2. dmbche ◴[19 Jul 25 17:37 UTC] No.44617526[source]▶
Which is not what any of the companies did

Anthropic ALSO get copyrighted material legally, but they pirated massive amounts first

replies(1): >>44618954 #
3. rpdillon ◴[19 Jul 25 20:14 UTC] No.44618954[source]▶
Apologies, I read your original statement as somehow concluding that you couldn't train an AI legally. I just wanted to make it extra clear that based on current legal precedent in the U.S., you absolutely can. Methodology matters, though.