Skip to main content

AI Agent Safety and Compliance Settings

Keep AI Agent conversations safe and compliant with opt-out detection, crisis guardrails, permissions and Quiet Hours.

A
Written by Aaron Sundukovskiy

Your Salesmsg AI Agent includes settings that keep conversations safe and compliant. It can detect opt-outs, respond with care to crisis language, skip opted-out or blocked contacts, follow your Quiet Hours, and limit who can manage AI Agents.


Contact opt-out detection

Your AI agents can now detect when a contact wants to stop receiving messages and handle it automatically. This is turned on by default for Unified Agent, so your agents stay compliant without any extra setup.

Turning it on or off

  1. Open your agent's Configure tab.

  2. Find the Contact opt-out detection toggle. It's on by default.

  3. Turn it off if you don't want the agent to handle opt-outs automatically.

When the toggle is on, an opt-out paragraph and a /opt-out action are added to your agent's prompt automatically. Turning the toggle off removes the /opt-out action.


Opted-out or blocked contacts

When you enroll contacts in an active AI Agent, opted-out and blocked contacts are handled like this:

  • Skip Opted-out or Blocked Contacts: Automatically skip these contacts during the enrollment process.

  • Disable Enroll Option for Opted-out/Blocked Contacts:

    • In the Contact Sidebar, if a contact is opted-out or blocked, the "Enroll in AI Agent" option will be disabled.

    • A message will appear: "Contact opted-out" or "Contact blocked."

    • Tooltips:

      • For opted-out contacts: "This contact has opted-out."

      • For blocked contacts: "This contact has been blocked.

  • Contacts Page Mass Action:

    • If any selected contact is opted-out or blocked, the "Enroll in AI Agent" option will be disabled.

    • A message will appear: "Contact(s) has opted-out or has been blocked."

    • Tooltip: "Contact(s) has opted-out or has been blocked."


Crisis and self-harm language guardrails

Your Texting AI Agent can recognize when a contact's message signals crisis or self-harm, and responds with care instead of continuing its normal script.

How it works

When a contact enrolled in a Texting AI Agent sends a message with crisis or self-harm language, the agent immediately stops whatever task it was running, whether that's a booking flow, an opt-out request, or anything else in progress. It doesn't finish the task first.

The agent replies with a supportive, pre-approved message and includes a crisis line number where appropriate. If the conversation is happening in a language other than English, this response is translated automatically.

The conversation is then handed off to your team. The agent creates an internal note summarizing the reason for the escalation and the time it happened, and tags the contact owner so your team can follow up.

Every crisis detection is logged separately for compliance and audit purposes, regardless of your other AI Agent settings.

Turning it on or off

You'll find Crisis/self-harm language detection as a toggle in your Unified Agent settings. It's turned on by default for every agent.

  • When it's on, the behavior above is active.

  • When it's off, your agent follows its standard conversation flow with no crisis-specific handling.

Turning this off does not change any of your agent's other instructions.


AI Agent permissions

Now in your settings there is a permission settings that allow users to enable or restrict access to the AI Agent feature.

This enhancement provides greater security and control for clients, ensuring that only authorized users can interact with or manage AI Agents. These permission controls help safeguard sensitive operations while allowing flexibility for different team roles.

You can change the access to AI Agent based on individual accounts or roles as a whole.


Quiet Hours

AI Agent messages, including auto-reminders and AI-generated first messages, respect your Quiet Hours settings.

If a message is scheduled to send during a restricted window, it will be held and automatically delivered at the next available sending time


Frequently Asked Questions

Is contact opt-out detection on by default?

Yes. The Contact opt-out detection toggle is on by default. Turn it off in your agent's Configure tab if you don't want the agent to handle opt-outs automatically.

Can I enroll an opted-out or blocked contact in an AI Agent?

No. Opted-out and blocked contacts are skipped during enrollment, and the Enroll in AI Agent option is disabled for them.

Is crisis and self-harm language detection on by default?

Yes. Crisis/self-harm language detection is turned on by default for every agent.

Do AI Agent messages follow my Quiet Hours?

Yes. A message scheduled during a restricted window is held and delivered automatically at the next available sending time.


Related Articles


Need Help? âœ‹đŸ»

Contact us on live chat or send an email to us at [email protected].

Did this answer your question?