Consent and disclosure

The two consent lines an agent speaks before its greeting, and the checks that stop a switched-on disclosure from saying nothing.

Before its normal greeting, the agent can speak two short consent lines in its own voice (see the opener in Voice and speech).

1  "This call is being recorded for quality and coaching purposes."   recording disclosure
2  "You're speaking with an AI assistant."                            AI disclosure
3  the agent's normal greeting

You can audition either line with the Preview button, which plays it in the agent's voice.

Both lines are per-agent overrides of a workspace default, like every other control on the Privacy page.

Recording disclosure

The recording disclosure tells the caller the call is being recorded. Turn this on whenever you record call audio.

This is the line that matters legally. In two-party-consent jurisdictions (Western Australia, for example) recording a private conversation needs the caller's consent.

When the disclosure resolves to on, it is spoken first, ahead of the AI disclosure and the greeting.

Preset lines

Pick one of the presets or write your own. The presets are plain, one sentence each, for example:

"This call is being recorded for quality and coaching purposes."
"Just so you know, this call may be recorded for training and quality."
"This call is being recorded."

AI disclosure

The AI disclosure tells the caller they are speaking with an AI assistant. This is best-practice rather than a legal requirement in Australia, and it is independent of recording.

It is spoken after the recording disclosure, before the greeting. Presets include lines like "You're speaking with an AI assistant."

Why an enabled-but-blank disclosure is blocked

A switched-on disclosure with no line blocks publishing

If you turn a disclosure on but leave its line empty, the agent would announce nothing: a silent consent failure. The editor rejects the save, and if the bad state somehow reaches publish, the readiness validator blocks it with an error.

Add the line the agent should speak, or switch the disclosure back off.

Three layers enforce this, so no path around the editor can slip a broken disclosure through:

At save -- the settings API rejects a switched-on disclosure with a blank line, whatever sent the change.
At publish -- the readiness validator raises a blocking error, so you cannot ship a version that discloses nothing. See the publish gate in Agents.
At call time -- the runtime speaks a disclosure only when it is both switched on and non-blank, and it never substitutes copy of its own.

An explicit "off" is honoured the same way, as a deliberate choice.

Risks you can acknowledge

The validator also flags two non-blocking risks:

recording is on but the disclosure is off
the disclosure is on but the agent does not store audio, so the spoken claim would be misleading