Tell people they are talking to an AI
The AI notice — an agent says it is an AI, and optionally that the conversation may be recorded, before anything else — and the workspace rules that require it and turn on redaction for new agents.
An agent can tell people, at the very start of every conversation, that they are talking to an AI, and that the conversation may be recorded. It is said before anything else, in the language the agent opens in, on phone calls, web calls and chats alike. Many regulations expect it; your own policy may too.
Turn it on for an agent
Open the agent, then Compliance. Under AI notice:
- Say it is an AI — the agent's first line begins with Just so you know, I'm an AI assistant.
- Say the conversation may be recorded — it adds This conversation may be recorded.
Each switch saves at once. What the agent says first then shows the exact opening line for each language the agent speaks: the notice, followed by the agent's own greeting. The notice is spoken in the language the agent opens the conversation in. An agent that waits for the other person to speak first is told to begin its first reply with the notice instead.
The notice is in the agent's language: Hindi, Bengali, Tamil, Telugu, Kannada, Malayalam, Marathi, Gujarati, Punjabi, Odia, Assamese, Urdu, French, Spanish, German, Italian, Portuguese, Dutch, Indonesian and Malay have their own wording; any other language uses English.
The agent's instructions also gain a line telling it that it is an AI and must never claim to be a person.
Off means untouched: an agent without the notice opens exactly as it always did. Nothing turns the notice on by itself unless your workspace requires it for new agents (below).
Workspace rules
Workspace settings → Agents holds two rules for every agent in the workspace. Anyone who can see agents can read them; only an Admin can change them. Both are off until an Admin turns them on.
- Every agent says it is an AI. New agents start with the AI notice on, and no one can turn it off on an agent while the rule is on. Agents that don't say it yet keep working; the setting shows how many there are, and you turn it on in each one's Compliance settings.
- Redact personal data in new agents. New agents start with personal-data redaction on: phone numbers, email addresses, ID and account numbers and similar details are removed from transcripts before they are stored. Existing agents keep their own setting.
Changing a rule never changes an agent that already exists, except that the AI notice can't be turned off while it is required.