Set your workspace's content tolerance

Last updated: October 26, 2026

Choose how much mature content model providers allow in your workspace, from strictest to most permissive, and lock it for every project.

Before you start

  • You need Manage workspace settings, which the Admin role has.

  • Content tolerance is set by model providers themselves. A model with no safety setting of its own is not affected by it; to screen requests to every model, use Input guardrails.

Steps

  1. Open Settings and select Content Policy under AI Controls. Under Content tolerance, the note shows the level the workspace inherits, for example Inherits level 6 from Layer's default.

  2. Select a level, from 1 (strictest) to 6 (most permissive).

  3. To stop projects and child workspaces from changing it, select the lock at the end of the row.

  4. Select Save.

Result

Every generation in the workspace asks model providers for the level you chose. Projects can set a stricter level of their own, never a more permissive one; child workspaces follow the level unless they set their own or you locked it. Layer's default is 6.

How the levels work at each level of control

  • Parent workspaces. A child workspace inherits its parent's level and can raise or lower it, unless the parent locks it.

  • Projects. A project can only be stricter than its workspace. Levels above the workspace's are unavailable in the project's AI controls, and a project's saved level above the workspace's does not apply. See Set AI controls for a project.

  • Groups and members. Content tolerance belongs to where work is made, so groups and members do not set it.

Production tips

  • Match the level to the game's rating. Start from the strictest level the game's age rating allows, and raise it only for projects that need it, such as a mature title's horror art, by setting a more permissive workspace level and stricter levels on the other projects.

  • Lock it when a publisher sets the bar. A lock keeps every project and child workspace at the level you chose.

  • Pair it with guardrails. The tolerance relies on each provider's own filter. Input guardrails screen the request before any model sees it, and work the same on every model.

If something goes wrong

  • The levels cannot be selected. A parent workspace locked the tolerance (Locked by the ... workspace). Ask an admin of that workspace.

  • A generation was refused or came back filtered. The provider applied the tolerance. See Your prompt or image was blocked.

  • Output looks the same at every level. The model has no safety setting of its own, so the tolerance does not reach it. Use input guardrails for that model.

Related