Guardrails & Brand Safety
Last updated: October 2026
Guardrails are rules every AI agent must follow when it writes for you. They are included in every generation request as absolute rules, and some are checked again right before a post is published.
Global prohibitions
Go to Settings → Guardrails → Global prohibitions and add anything your content must never include, such as competitor names, pricing claims, or medical claims. These apply to every agent and every brand in your workspace.
Blackout periods
Under Settings → Guardrails → Blackout periods, add a name and a start and end date (for example “Election Day” or “Q4 earnings”). While a blackout is active, agents won't generate content, and any post due to publish during it is vetoed instead of going out. The post stays in your workspace with the blackout noted as the reason, so you can reschedule it afterwards.
Built-in rules
These apply to every workspace and cannot be switched off:
- No naming specific competitor brands
- No profanity, slurs, or offensive language
- No unverified claims about market data or statistics
- Nothing that could read as financial advice
- Nothing that breaks the target platform's terms of service
Per-agent rules
On an agent's page under Guardrail overrides, you can add stricter rules for that agent only: block politics, block adult content, block any competitor reference, require a disclaimer, or add a custom word/phrase blocklist. Agent rules can only add restrictions — they can never loosen your workspace rules.
Brand-level rules
Each brand's voice settings include banned words and forbidden visual elements. The copywriter avoids those words and the image generator avoids those visual elements.
Before publishing
Every post gets a content-safety check right before it publishes. Posts still go through your Review Queue first.
Related articles
Still stuck? Email support