Does Palette train on your data?

Their words · Terms of Service

“ The services use artificial intelligence, including third-party generative models, to produce video, images, audio, characters, and related media from your prompts and reference material. Generative output is probabilistic: results vary between runs, may not match your prompt, and may contain visual, factual, or textual errors. Review Outputs before publishing or relying on them. We process prompts, reference material, and related context only to generate Outputs and operate the service, and we apply the same confidentiality and security commitments to that processing as to other Customer Data. We do not use Customer Data or Outputs to train any model, whether our own or a third party's, and our agreements with model providers prohibit them from doing so. Model availability changes over time. We may add, substitute, or retire the underlying models we use, and results from the same prompt may differ as a result. Where a feature depends on a specific model, we will give reasonable notice of a material change where practical.”

Captured 2026-09-25. Open captured source →Snapshot SHA-256: 4a14de31997204f1e232bccb4a6ada4283aa2e21f699c43675410a3d1e5fe754