Skip to content
FastGen Cloud

Content policy

FastGen Cloud is a general-purpose image and video API. Prompts and negative prompts are passed to the model exactly as you send them — no injected terms, no hidden prefixes, no “style steering”. What your product allows is up to you and the Acceptable Use Policy. One thing is enforced automatically and cannot be turned off: prompts for sexual content involving minors are rejected. Everything else — prompts, input images and generated output — is passed through and generated as-is.

Enforced automatically

Sexual, suggestive or exploitative depictions of minors, and attempts to produce them. The prompt is matched against a deterministic, local minor-safety term list before the job is queued. This runs in-process: there is no third-party classifier in the path, no network call, and no copy of your prompt leaves the platform to be scored.

The gate reads the positive prompt. The negative prompt is deliberately not scanned for these terms, because age and explicitness terms in a negative prompt are a safety measure rather than a request.

Not checked

Listed explicitly, because assuming a safety net that is not there is an expensive mistake to make in production. None of the following happens:

  • General prompt filtering. There is no adult, violence, or topic keyword list. Adult content between adults passes untouched.
  • Input image scanning. Edit sources, img2img references, Face ID photos and video start frames are not classified.
  • Output scanning. Generated images are not classified before the URL is released, and video frames are not sampled and scored. A job never fails for what it produced.
  • Prompt rewriting. No language model sits in front of the sampler. The only strings prepended are a checkpoint’s native quality tags and a catalog LoRA’s trigger word when you select that LoRA.
  • Provenance metadata. No C2PA or equivalent marking is embedded in output.

If your product needs any of these, you implement them — moderating user-generated images covers where to put the gate.

Policy, enforced on report

The AUP also prohibits non-consensual likenesses of real people, illegal content, targeted harassment and hate, and deceptive media presented as authentic. These are handled on report and review (content removal, suspension, termination), not by automatic blocking. If you expose generation to end users you are the publisher: age-gate adult features, keep consent for real people, and label generated media. See age-gating a generation product.

What a block looks like

Blocks are synchronous and scoped to the request. The call returns 422 before the job is queued, so nothing is charged and there is nothing to refund.

422 — prompt gate
HTTP/1.1 422 Unprocessable Entity

{
  "error": {
    "type": "content_blocked",
    "code": "content_blocked_prompt",
    "message": "Request blocked by content policy (prompt). See https://fastgencloud.com/legal/aup",
    "request_id": "req_…"
  }
}

moderation.status on a job reflects this prompt gate only. It is not a content classification of your image and should not be used as one in your own product.

No automatic account enforcement

A block records a moderation_events row for audit and notifies an administrator. It does not suspend the account, pause a key, revoke credentials or issue an automatic penalty — a blocked request costs you that one request and nothing else. Enforcement under the AUP is a human decision made on review, not a threshold your integration can trip by accident.

Payment fraud is the exception and is unrelated to content: a chargeback suspends the account and revokes its keys.

Building for the policy

  • Treat content_blocked as a user-facing validation error, not a transient failure — do not auto-retry the same prompt.
  • If your product accepts free-text prompts from end users, surface the block to them in your own words and keep your own rate limits.
  • The model catalog is curated; the API never accepts arbitrary checkpoints, LoRA files or raw workflow graphs.
  • Outputs are deleted after 7 days. Copy anything your product needs to keep into your own storage.

False positives

The term list is deterministic, so it occasionally catches innocuous phrasing — an age word used as an adjective, or an adult-only disclaimer written into the prompt itself. If you believe a request was blocked in error, email support@fastgencloud.com with the request_id and we can review the match.