Acceptable Use & Synthetic Media Policy

Prohibited uses

Do not generate or edit likenesses of celebrities or public figures, including politicians and deceased famous people. Deceptive impersonation, fraudulent endorsements, sexual exploitation and nudification are prohibited. Do not bypass safety controls.

Available generation

Ordinary private-person portraits, fictional people, landscapes, products and animals are allowed. A photorealistic face alone is not a reason for rejection. Supported reference images, video continuation and audio are reviewed along with prompts and generated media.

Review and enforcement

Prompts and generated media are automatically reviewed before delivery, using a review model through Replicate. Requests are stopped if review is unavailable, fails, or is uncertain. A rejected input does not consume credits; failed output review follows our failed-generation credit return process. Automated review can make mistakes and does not guarantee detection of every violation.

Five policy-denied reviews under the current policy within 24 hours temporarily restrict further generation until the rolling count falls below five. We may investigate reports, restrict accounts, and remove violating content. Previously generated content without a recorded review is withheld from the creation gallery and task responses; previously downloaded copies cannot be recalled.

Data used for review

The original prompt and generated media are sent to Replicate and its review model provider for safety classification. Safety audit records store the account identifier, time, review stage, decision, reason category, policy version and associated task identifier, without duplicating submitted prompts or media. Audit records are removed after 30 days when further reviews run. Normal generation history is stored separately under our privacy policy.

Report abuse or appeal

Email support@flux3.dev with the task identifier or page URL, the reason for your report and a contact address. You do not need an account to report abuse. Do not attach sensitive identity documents or intimate imagery. You may use the same address to appeal a blocked request.