Settings and activity
1 result found
-
16,628 votes
An error occurred while saving the comment 
its_me_2048
supported this idea
·
Translate Ideas and Comments
Choose language:
There was an error during translation
Mike Minton's statement ("Our Terms of Service already permit training of models") is misleading: the ToS do not specifically address GenAI training but only the IP sublicensing of the creator content, not of personal data (such as the face, voice, etc. of the streamer).
As a matter of facts, the Privacy Notice explicitly mentions GenAI training, but:
1) this clause is not GDPR-compliant as it doesn't state the legal basis for the GenAI training (and that Minton publicly admitted himself he has no clue of the actual scope nor purpose for this GenAI training); this means that any GenAI training on an EU data subject would be unlawful as it would be incompatible with the GDPR; moreover, regardless of the current status (enabled or not) of the "Training for Generative AI" toggle, any past GenAI training of content containing the face and/or the voice of the streamer constitutes a violation of the GDPR;
2) the Twitch Privacy Notice does only apply to the streamer themselves, and not any other person (such as guests in a Discord calls or people appearing on broadcasted videos, including IRL bystanders) whose voice or image would appear on the stream produced by the creator; any GenAI training on such content without their own prior explicit and informed consent (which cannot be covered by the Twitch Privacy Notice) would itself constitute a violation of the GDPR;
3) the very existence of this "Training for Generative AI" toggle, and the fact that is it enabled by default means that the streamer is (as long as this setting is enabled) legally responsible towards Twitch to ensure to collect the prior explicit and informed consent of every single person visible or heard on their stream to have their personal data exploited for GenAI training (along with its exact purpose); moreover, if any audio or visual assets (would it be a part of an overlay, an emote, a background picture, a video or video excerpt, would it be a single frame, a soundtrack, a video-game or any audio or visual asset of a video-game) explicitly forbids GenAI training and Amazon actually does perform GenAI training on this content, then, in addition to exposing the legal responsibility of Amazon towards the original asset's owner, it would also engage the legal liability of the streamer towards Twitch.
I would strongly and urgently recommend Twitch to review their position regarding GenAI training on their user's content as well as the wording of their ToS and Privacy Notice (and how these address personal data of third parties who are not Twitch account holders but may appear or be heard on broadcast content) with a legal team familiar with international (and more specifically: EU) regulations as well as with representatives (such as Twitch Ambassadors) familiar with the actual type of content broadcasted on Twitch.