Reply to thread

Message: [QUOTE="Shamiso, post: 92748, member: 160"] Suno sets newly created songs to Link Only by default, but its current terms do not describe private visibility as a training opt-out. A private song stays off your public profile and the home page, while anyone with the direct link can still access it. The important split is between who can see a song and what Suno says it can do with Content behind the scenes. Suno’s privacy controls deal with exposure to other users, while its Terms of Service and Privacy Notice separately describe model improvement, data use, and retention. Suno’s own California disclosure makes the distinction harder to ignore. It says certain post-training datasets may include user Content and User Activity Information, which means the relevant question is not simply whether a track has a Public badge. [HEADING=2]Private visibility controls who can find the song[/HEADING] Suno’s help material describes Link Only as the default state for newly created tracks. Those songs are not displayed publicly on your profile or surfaced on the Suno home page, and changing visibility to Public makes them eligible for wider discovery. The wording is narrower than saying every private generation becomes raw base-training audio. Suno calls these post-training datasets but does not identify which individual songs entered them, which model-development stage they served, or whether every user track was eligible. Still, the disclosure confirms user Content can appear in at least some model-development datasets, making visibility status a poor proxy for internal use. The current terms use similar language around privacy settings. They say users can change settings so Output remains private instead of being publicly available through certain third-party sharing defaults. Nothing in that passage says private Output is excluded from product improvement or AI model development. The broader Content license sits elsewhere. Suno says using the service or transmitting Submissions grants it a worldwide, non-exclusive, fully paid-up, sublicensable, assignable, royalty-free, perpetual, and irrevocable license over Content for purposes that include improving its products, services, and related artificial intelligence and machine-learning models. This is where [B][URL='https://goldmidi.com/community/threads/suno-has-said-v6-was-trained-on-user-creations.78133/']community creations behind Suno v6[/URL][/B] become relevant. Suno has said community creations were among the inputs involved in developing v6, so privacy settings and model-development permissions are not interchangeable concepts. [HEADING=2]Suno’s privacy notice goes further than the song toggle[/HEADING] Suno’s June 2026 Privacy Notice says certain User Activity Information, Submissions, Interactive Chat Information, and other Content can be used to improve its services, including training and enhancing the models that power them. The notice does not limit that wording to tracks marked Public. It also separates personalization controls from broader model improvement. Users may be able to view, edit, or disable a personalization profile where Suno provides one, but the document describes training and service improvement under a different section. Turning off a taste profile therefore should not be read as a general Suno training opt-out unless Suno explicitly says so. The same contract split showed up in the previously published [URL='https://goldmidi.com/community/threads/suno-paid-song-ownership-still-leaves-a-training-license.78160/']paid ownership and training-license article[/URL]. Paid users can receive Suno’s ownership interest in qualifying Output while Suno still reserves a separate Content license, which is another reason a privacy switch should not be assumed to revoke permissions written elsewhere. [HEADING=2]Deleting information is a separate problem again[/HEADING] Suno’s Privacy Notice lets users request deletion of certain information, subject to exceptions. It also says Suno may retain information for contracts, legal proceedings, compliance, auditing, or where data has been de-identified, anonymized, or aggregated so it no longer identifies the person. A deletion request is therefore a different tool from making a song private. It also should not be casually described as wiping a song out of an already trained model, because removing stored source material and removing learned influence from a trained system are technically different jobs. [B][URL='https://link.springer.com/article/10.1007/s10994-022-06178-9']Machine unlearning techniques[/URL][/B] exist because deleting training data from a model can require more than deleting the original file. Suno’s current public documents do not give ordinary song visibility controls a label such as “do not train,” “exclude from model improvement,” or “training opt-out.” Its [URL='https://goldmidi.com/community/threads/sunos-upload-terms-became-evidence-against-it.77662/']upload-terms language[/URL] instead places broad downstream permissions around submitted material, including model training and improvement. So setting a track to Link Only can still be useful if you do not want it discoverable across Suno. It changes audience access. Based on the current documents, it does not by itself amount to a stated instruction that Suno must exclude the track or associated Content from model development. [/QUOTE]

Name