AI Protection Hub

There is no general artist opt-out that removes music from Suno's past model training. Major record labels sued Suno over alleged unlicensed training on copyrighted recordings, and Suno has publicly defended training on music available on the open internet as lawful. Warner Music later reached a licensing partnership with Suno, but that does not create a universal opt-out for independent artists.

There is no public, general opt-out that removes music from Udio's past training. Udio was sued by major labels over alleged unlicensed training; some settlement/licensing reporting indicates parts of the industry are moving toward authorized catalogs, but independent artists should not assume their works are excluded.

Stability AI has account-level controls for whether some user content helps improve its services, and Stable Audio has described using licensed datasets for model training. This is not the same thing as removing your already-published music from every AI dataset, but it is one of the clearer controls for content you provide directly to Stability products.

YouTube offers controls for whether third-party companies may use your YouTube content to train AI models. That setting should not be described as a full opt-out from Google's own AI systems or from uses already covered by YouTube's terms. For musicians, the most accurate action is to use the third-party control, manage visibility, and keep rights records ready for copyright complaints.

Spotify has no artist-facing opt-out that says 'do not use my catalog for any AI purpose.' The clearest actions are to review Spotify's rules, keep your distributor agreement tight, use required AI disclosure metadata when applicable, and report impersonation or infringing uploads. Avoid claiming Spotify categorically never uses catalog data for AI unless you have a current written policy from Spotify.

Apple provides information about Applebot and data used for Apple Intelligence, including ways for websites to control crawling. That is different from an Apple Music artist catalog opt-out. For musicians, the practical step is to keep distribution agreements clear and use Apple's published web-crawling controls for websites you operate.

SoundCloud's terms include broad language about AI technologies, but SoundCloud has publicly said it has not used artist content to train generative AI models and does not allow third parties to scrape SoundCloud content for AI training. There is no dedicated artist opt-out portal today; set tracks private and monitor policy changes.

Meta says it uses public Facebook and Instagram content to train generative AI and does not use private messages for that purpose. EU/UK users have objection rights and access to Meta's objection flow. In the US and many other regions, there is no general opt-out for public posts; making accounts private is the clearest risk-reduction step.

X provides a Grok data-sharing control that can limit future use of your posts and interactions for Grok training/improvement. It is one of the clearer global controls, but it does not undo any past training or stop copying by third parties who already accessed public posts.

TikTok's public content and broad platform license create exposure for short clips, previews, lyrics, and artist likeness. There is no simple global 'do not train AI on my posts' control. EU users may have data protection objection rights, but musicians should still reduce public exposure for unreleased or sensitive material.

ChatGPT users can opt out of having chats used to improve OpenAI's models. That protects future ChatGPT conversations, prompts, uploaded files, and similar account content depending on plan and settings. It is not a catalog-wide opt-out for music already published on the web or licensed through third parties.

ElevenLabs is mainly a voice AI risk for musicians: unauthorized voice clones, spoken/sung likeness, and synthetic endorsements. The practical control is to report unauthorized voice clones and avoid uploading voice samples unless you understand the terms. This is different from opting a recording catalog out of music-model training.

Bandcamp is generally lower risk because it is a direct-to-fan store rather than an AI company, and it has not publicly announced a generative-AI training program for artist catalogs. But there is no specific AI-training opt-out portal. Use Bandcamp as part of a direct-sales strategy, and keep monitoring terms after ownership or policy changes.

Distributors are a key control point because they deliver your recordings, metadata, and rights data to platforms. Do not assume all distributors have the same AI policy. Read your own agreement and ask directly whether your catalog, metadata, or analytics can be licensed or shared for AI training.

Unchained Music markets itself as artist-protective around AI, but artists should verify the current written policy and distribution agreement before switching. Treat it as a distributor to evaluate, not a guaranteed universal solution.