Your prompt or image was blocked

Last updated: October 12, 2026

A generation can be refused before it runs, or come back with nothing, when its prompt, its input images or the style applied to it fail a content check. This page explains which check stopped it and what to change.

How to tell which check stopped it

The failed tile on the board and the viewer both show the reason. The wording tells you where the refusal came from:

  • "This request was blocked by your workspace's ... guardrail." One of your workspace's own input guardrails refused it. The message names the guardrail and usually says why.
  • "Rejected by our content safety system." The model's provider refused the request or its output as unsafe content.
  • "Your prompt may contain intellectual property references that cannot be processed." The provider refused a reference to a protected character, brand or work.
  • "The model returned no result." The model sent nothing back and gave no reason, which in practice usually means it declined the inputs.

A refusal is shown as a warning, not an error: nothing in Layer is broken, and the request needs to change.

Most likely causes

A workspace guardrail blocked it

Workspace admins can set input guardrails in Settings > AI Controls > Input Guardrails, and people who manage a project can add more for that project under the project's Content policy. Before a generation runs, Layer shows each guardrail the generation's prompt, its input images and the names of any styles or reference sets applied to it, and asks whether the request breaks the rule. Layer offers three built-in guardrails, and admins can write their own:

  • Copyright infringement: reproducing a specific copyrighted work, such as named films, games, series or their assets, whether named in the prompt or supplied as an input image.
  • IP & trademark infringement: a real brand's logo, mascot, packaging or trademarked characters, including close imitations.
  • Artist style mimicry: imitating a named living artist, in the prompt or through a style or reference set named after them.

A guardrail set to block refuses the request, and nothing is sent to the model. A guardrail set to warn lets the generation run and records the match; the viewer lists it under Generated despite in the file's details.

What to do:

  • Read the guardrail's name and reason in the message, then rewrite the prompt without the reference it names. Describe the qualities you want (a genre, a palette, a medium, an era) instead of naming a franchise, brand or artist.
  • Check the input images and the applied style or reference set too. A neutral prompt is still blocked when an input image is someone else's artwork to copy, or when a style's name identifies an outside artist.
  • If the request is legitimate (your own game, a brand you are licensed to use, an artist you commission), ask a workspace admin to add it as an exception to that guardrail. Built-in guardrails take exceptions that narrow what they flag.
  • Running the same request again gives the same answer, so change the request before you retry.

The model's provider refused it

Each model's provider runs its own safety checks, and some refuse content that another model accepts. Your workspace's Content tolerance in Settings > AI Controls > Content Policy sets how strict those checks are, from Strictest to Most permissive, where a model lets Layer choose. A project can be stricter than its workspace, never looser.

What to do:

  • Rephrase the prompt to remove the violent, explicit or otherwise sensitive wording the provider is likely to flag.
  • For an intellectual property refusal, remove named characters, brands and titles, as above.
  • Try another model for the same job. Providers draw the line in different places.
  • If your studio's work needs mature content, a workspace admin can review the content tolerance setting.

The model returned nothing

When a model returns no result without a reason, treat it as a soft refusal of the prompt or inputs. Rephrase the prompt, swap any input image that might be the trigger, and run it again. If it keeps happening with harmless inputs, see Fix a failed or stuck generation.

Still stuck?

Contact support from Help & Resources > Talk To Us in Layer, and include the session link, the time of the run and the exact message on the failed tile. If you think a workspace guardrail flagged a request by mistake, tell your workspace admin as well: they can change its exceptions or switch it from block to warn.

Related