Guardrails & Brand Safety
Last updated: June 2026
Guardrails are rules you configure to prevent AI-generated content from going in directions your brand should never go. They run automatically on every post before it enters the review queue, so problematic content is caught before a human even sees it.
Prohibited topics
Add topics, keywords, or phrases that should never appear in your content. Go to Settings → Brands → [Brand] → Guardrails → Prohibited topics and enter each item. Examples:
- “competitor names” — prevents Anthyx from mentioning rival brands
- “political commentary” — keeps content neutral on sensitive topics
- “price guarantees” — avoids legal risk from unverified pricing claims
- “medical claims” — required for health and wellness brands
The Guardrails agent performs a semantic check — it catches paraphrased versions of prohibited topics, not just exact keyword matches. A post that discusses “the brand that rhymes with Nike” when Nike is prohibited will still be flagged.
Tone limits
Beyond the tone slider in brand settings, you can set hard tone limits under Guardrails → Tone rules. Examples:
- Never use exclamation marks more than once per post
- Avoid superlatives (best, greatest, #1) unless citing a third-party source
- Do not use first-person singular (I, me, my) — brand voice is always “we”
Tone rule violations cause the Guardrails agent to automatically request a revision from the Writer agent before the post surfaces in your review queue.
Publishing blackout windows
Blackout windows prevent Anthyx from scheduling posts during specific time periods — for example, during a crisis, a board meeting, earnings season, or a public holiday. Configure them under Settings → Brands → [Brand] → Guardrails → Blackout windows.
You can set recurring blackouts (e.g. every Sunday, or the last week of each quarter) or one-off windows with a specific start and end time. Posts that are approved and would fall inside a blackout window are held in the queue and released at the next available time after the window closes.
What happens when a guardrail triggers
When the Guardrails agent flags a post, it does not silently modify it. Instead:
- The post is held in the queue with a Guardrail flag badge.
- The specific rule that was violated and the offending passage are shown in the post detail view.
- You can dismiss the flag (override the guardrail for this post) or send the post back for revision.
- All guardrail triggers are logged under Settings → Guardrails → Audit log for compliance review.
Related articles
Still stuck? Email support