Your social media AI training opt-out options are, in most cases, either non-existent, buried three menus deep, or conditional on living in the right country. Twitch made waves recently by announcing it would let users opt out of having their streams, VODs, and chats used for AI training. Turns out, by the standards of the industry, that makes Twitch one of the more reasonable players in the room.
Almost every major social media platform is training some form of AI on your data. Not one of them operates an opt-in model. If you logged on today, the default assumption, across the board, is that you consented.
Meta, Facebook, Instagram and Threads: The Broadest Reach
Facebook’s parent company Meta has perhaps the most expansive data appetite of the lot. Its policy covers your posts, photos, interactions, data from third-party brokers, and, in its own words, “information that is available on the internet.” Meta says it stops short of training on private messages unless someone in the chat shares those messages with its AI tools. For most users outside the EU, there is no opt-out.
The EU situation has its own recent history worth knowing. In June 2024, Meta informed the Data Protection Commission of its plans to pause training its AI model on European users’ data. That pause did not last. According to REVERA, Meta actually commenced using EU users’ public data on 27 May 2025. The window to object mattered: according to KP Partners, users who submitted a formal objection before 26 May 2025 effectively prevented their publicly available content from being included in AI training. That deadline has now passed. KP Partners also notes that on 23 May 2025, the Higher Regional Court of Cologne preliminarily affirmed that Meta’s legitimate interest in using that data outweighs affected users’ rights to informational self-determination, complicating any future legal challenge.
For those outside the EU, your practical options with Facebook are limited. You can file a complaint if your personal information appears in an AI-generated response, which Facebook will then review at its discretion. Interacting with Facebook’s AI tools to check whether your data has been scooped up is itself treated as consent to train future models. The only reliable move is not to engage, though deleting your account won’t erase data Facebook already holds.
Instagram, owned by Meta, operates under the same framework. It briefly ran a feature called Muse that let users generate AI images of other users without their consent, before removing it. The opt-out position is identical to Facebook’s. Threads, also Meta-owned, follows the same rules: publicly available data is used for training regardless of consent, unless you are in the EU and submitted your objection before the deadline.
Platform by Platform: What You Can Actually Do
LinkedIn collects posts, comments, profile data, resumes and group activity for AI training. It was also reportedly sued over allegations it trained AI on private direct messages, which LinkedIn denies. The site at least gives you a toggle. In Settings and Privacy, go to Data Privacy, then How LinkedIn uses your data, then Data for Generative AI Improvement. Switch off “Use my data for training content creation AI models.” Per LinkedIn’s official policy, this only covers data collected on LinkedIn, not the wider Microsoft ecosystem.
Snapchat uses public images, video and audio to train generative AI for features like AI Snaps and AI Lenses. The opt-out is relatively straightforward: in Settings, under Privacy Controls, open Generative AI Settings and disable “Allow Use of Public Content.” Snapchat’s design makes third-party scraping harder than most, though not impossible.
X trains its Grok AI on public posts and, following the company’s merger with xAI and SpaceX, that data can be used across the wider company. You can opt out via Settings, Privacy and Safety, Data sharing and personalisation, then Grok and Third-party Collaborators. Uncheck the relevant boxes, including the one covering your interactions with Grok and xAI.
Reddit does not train its own AI models but has deals with both OpenAI and Google to train theirs on Reddit data. The company has reportedly considered ending those arrangements. There is no opt-out tool. Deleting your account is the only practical option, and even that won’t reach data that was scraped before you left.
TikTok‘s structure became more complicated after its US operations were spun off into a separate entity. ByteDance trains AI models on user data, though US-based users’ data is reportedly restricted from use in ByteDance’s broader AI model training. Opting out requires filing a specific privacy breach objection under local law, not a simple toggle. There have been reports that TikTok trains even on private videos and unsaved drafts, making account deletion the most reliable protection.
YouTube is tied to your broader Google account, which makes the privacy picture considerably more complicated. Google has been found to train video-generation models on YouTube uploads without creators’ knowledge. Regular users can manage some settings at Google account level. Creators uploading videos currently have no option to prevent Google from training its own models on their content, though they can opt out of third-party AI training via YouTube Creator Studio under Settings, Channel, Advanced Settings, then Third-party training.
Twitch allows its policies to cover streams, VODs, clips, chats and channel imagery for generative AI training. To opt out, go to your Twitch security settings, scroll to Training for Generative AI, and toggle it off. The setting covers your channel only: data you generate on someone else’s stream is governed by that streamer’s own preferences, not yours.
For EU users with Meta products specifically, the 26 May 2025 objection deadline has passed, and the legal landscape following the Cologne court’s May ruling means future challenges face a harder road. Everywhere else, the default remains opt-in by presence, opt-out by effort, and no guarantees.
