Guardrails and rules
Open Controls in the sidebar of your org Space. The page shows what Empathy.AI always applies, then the rules that belong to this Space.
Platform guardrails
The Platform guardrails card is the floor.
- Set by Empathy.AI
- Always-on, including immutable safety rules
- Read-only — you cannot edit, disable, or delete these
Guardrails always apply. They are not optional safety settings.
Rules on the card appear as chips. Locked chips are safety rules; others are platform policy. Extra chips collapse behind + N more.
Space rules
Each rule is a card with a name, a status badge, and metadata (when it was updated, who updated it, and a version). The body is rendered by type — faders, format badges, a list, or plain text.
| Status | Meaning | What you can do |
|---|---|---|
| Active | Live in the Space | Disable |
| Inactive | Exists but not applied | Enable |
| Draft | Not published | Delete |
Use the overflow menu (⋯) on a card:
- Disable an Active rule.
- Enable an Inactive rule.
- Delete a Draft only. Confirm with Delete draft rule? — the draft is removed because it was never published.
Published rules are not deleted from this menu. Disable them instead.
How rules appear
Tone & voice. Equaliser-style faders for Formality, Directness, Warmth, Verbosity, and Technical depth, plus body text and an optional preview quote.
Response format. Badges for bullet lists on long procedures, including document titles, default answer length (Short / Medium / Long), and mirroring the question language.
Domain context. A list body — the first items, then + N more.
Plain. Some rules show as body text.