Post B8xc5btBv5FPD266mu by chris@mastodon.chrispelli.fun
 (DIR) More posts by chris@mastodon.chrispelli.fun
 (DIR) Post #B8xSU35XRjw2ySt53g by EUCommission@ec.social-network.europa.eu
       0 likes, 2 repeats
       
       From now on, companies across the EU must be clear about  their use of AI.Providers and deployers of AI must:✔️ AI-generated content must carry a machine-readable mark✔️ Deepfakes and AI text on public interest topics must be visibly labelled✔️ When you chat with a bot or virtual agent, it must make clear it’s not a human✔️ When AI is actively analysing you, companies must tell you️We've published guidance to help companies comply.Transparency helps build trust 👇 https://link.europa.eu/gTvhRX
       
 (DIR) Post #B8xXZ8DoHsuI7801aa by davidrevoy@framapiaf.org
       0 likes, 0 repeats
       
       @EUCommission Dear European Commission,I appreciate the initiative, but I have concerns. Will the platform use AI to detect AI-generated content? As an artist whose work was trained on without consent for 20 years, detection tools now incorrectly flag my human-made art as "likely AI-generated" ( see: https://framapiaf.org/@davidrevoy/116941001377913004 ).What safeguards exist to prevent this?RT: https://framapiaf.org/users/davidrevoy/statuses/116941001377913004
       
 (DIR) Post #B8xYRPf7ezB7qHeqq8 by phoerious@mastodon.social
       0 likes, 0 repeats
       
       @davidrevoy @EUCommission The platforms will most likely use watermarks in AI content. If calibrated correctly, they have very low false positive rates. However, they are also easy to remove and it is unclear what platforms will do with AI content generated with unwatermarked Open-Weight-Models etc.
       
 (DIR) Post #B8xZPqNHTQJC4180kC by chris@mastodon.chrispelli.fun
       0 likes, 0 repeats
       
       @davidrevoy @EUCommission I think this is targeted at those who provide services to generate AI content. They have to add watermarks to the generated content.
       
 (DIR) Post #B8xb5cJAGJzCYQgZai by davidrevoy@framapiaf.org
       0 likes, 0 repeats
       
       @chris @EUCommission Thank you. I see. I will read the details again.In any case, if it is limited to forcing AI provider to watermark in UE, a workaround sounds very straightforward (even to a 12 years/old): a user could simply purchase a VPN, then generate AI in a region outside the EU without the imposed watermark, and then post back the result on EU networks without being flagged as AI.
       
 (DIR) Post #B8xc5btBv5FPD266mu by chris@mastodon.chrispelli.fun
       0 likes, 0 repeats
       
       @davidrevoy @EUCommission I am not sure how it works. I didn't read the legal document but one avenue would be to require a company that operates in the EU to always mark the posts. But I don't know if the law is written that way.Google already rolled out their watermarking but that is only in metadata. I think the EU demands that the watermark is visible.
       
 (DIR) Post #B8xcqpPJTf78kGdaj2 by sgued@pouet.chapril.org
       0 likes, 0 repeats
       
       @davidrevoy @chris @EUCommission la loi va forcer les réseaux sociaux à détecter les posts IA? C'est pas l'impression que j'ai. J'ai juste lu l'article du monde sur le sujet et j'ai plutôt l'impression que ça va être les créateurs eux-mêmes (sous réserve qu'ils soient professionnels) qui vont être forcés à le faire.
       
 (DIR) Post #B8xobRUDKcdPf6MN5U by mindaugas@mastodon.0011.lt
       0 likes, 0 repeats
       
       @davidrevoy related story https://youtube.com/watch?v=OX1yHmeLPy8 (video "The Kurzgesagt Situation is Insane")@EUCommission