Higgsfield (consumer/creator tiers): you keep Outputs and may use them commercially, but inputs and outputs may train Higgsfield models, outputs may not train any other model, and the underlying model can be swapped without notice

Primary Higgsfield Terms of Use (last updated 26 Jul 2026). Ownership: Company "does not claim ownership" of Inputs or Outputs and does not restrict commercial use; output rights survive cancellation or account deletion and can be sublicensed to clients (s4.4). Training: Your Content, Inputs and Outputs "may be used by Company to train" its and affiliates' models; opt-out is only by deleting content or the account, going forward; Enterprise Agreement customers are excluded from training and treated as confidential (s4.4). Marketing reuse only for content you made public on the service or with consent. Likeness: uploading a person's name/likeness/voice means you represent you hold all releases and consents, including for generating Outputs (s4.2(b)). Outbound restriction: you may not use the Service, Outputs or derived data to train, fine-tune or distill any machine-learning model (s5.2(iv)) - this matters for pipelines that use generated plates as LoRA/fine-tune data. Model churn: Higgsfield may substitute, deprecate or retire any model at any time; a substitution "does not, by itself" breach the service; pro-rata refund only for material degradation of a paid plan (s1.7). Storage: retention varies by tier; content may move to cold storage, and after 30 days with no login, generation or subscription an account can be classified inactive (s3.4). Disclosure: user must disclose AI generation where the law requires and not strip provenance; Higgsfield may embed C2PA/invisible watermarks but does not warrant they persist (s5.5, s6.4). Practical reading: compared with enterprise aggregator terms (Runway Enterprise C3-B471-01: no-training commitments, indemnity covering third-party models) the creator-tier aggregator terms give output ownership but no indemnity statement was found in these sections, permit training on your uploads, and make reproducibility depend on the platform keeping a given model - record the model/version actually used per shot and keep local exports. Curator correction 2026-10-24 (discovery audit record C3-B1096-01, curator-reviewed C3R-0701; excerpts re-located by independent curator GET 2026-10-24): Correction. C3-B472-01 states the outbound restriction in section 5.2(iv) as a bar on using the Service, Outputs or derived data to 'train, fine-tune, distill' any machine-learning model, and dates the terms 'last updated 26 Jul 2026'. The clause as published on 2026-07-30 (archived copy, 'Last updated: July 26, 2026') and as live on 2026-10-24 ('Last updated: October 23, 2026') continues with two exceptions that the record omits: (a) models you create, train or fine-tune using the Service with your own Inputs, but not with Outputs of the company's AI models, and (b) uses expressly authorised in writing by the company. So the record's practical reading for Output-based LoRA or fine-tune data still holds (Outputs stay barred without written authorisation), but a model trained through the Service only on your own Inputs falls within the exception to s5.2(iv) (inference: an exception to that clause only; other terms are not analysed here), and the record's wording was broader than the clause already when written. Dating: the live page is now dated 23 Oct 2026; a comparison of the 30 Jul 2026 archived text with the live text shows differences in the removal-request, vulnerability-disclosure and agreement-update clauses; the sections the record relies on (1.7 model changes, 3.4 inactive accounts, 4.4 ownership and training, 5.2(iv), 6.4 marking) read the same in both copies as far as text-compared. The record was accurate as to ownership, training-on-content, model-substitution and provenance wording; only the omission of the 5.2(iv) exceptions is corrected, and the terms date is superseded. Curator qualifications: All cited documentation excerpts located on the live pages by curator GET; no video sources. Selected sections of the terms read closely; indemnity and limitation sections not read. Not legal advice. Limitations: Terms read once via rendered page; only sections 1.7, 3.4, 4.2-4.4, 5.2, 5.5, 6.4 read closely - indemnity/limitation sections (13-14) not reviewed, so absence of indemnity is not established. Enterprise Agreement and Privacy Policy biometric section not read. Interaction between s4.4 commercial use and s5.2(iv) for client deliverables used as training data by the client is unclear. Not legal advice. Evidence status: statements located in the cited sources by a correlated desk review (Claude); not executed; not independently reviewed.

{
  "citations": [
    {
      "url": "https://higgsfield.ai/terms-of-use-agreement",
      "note": "s4.4 User Inputs and Outputs; s4.4 User Inputs and Outputs; s5.2(iv); s1.7 Changes to Models and Features; s6.4 Provenance and Watermarking"
    }
  ]
}