Call My Agent is not an emergency service. Your AI phone agent won’t dispatch police, fire, or medical responders. If you or your caller is in danger, hang up this line and call 911 from another phone.
When the AI agent detects emergency intent on a call — words like “fire,” “ambulance,” “police,” “I’m being attacked,” “heart attack,” “overdose,” and similar — the agent immediately:
If the agent’s emergency-intent detection fails for any reason and the caller indicates an emergency in some other way (we miss it on the first pass), the call is still recorded and transcribed so the account holder can review it and follow up.
By using Call My Agent, you (the account holder) agree to all of the following:
This is also written into our Terms of Service (section 11). You agreed to it when you signed up.
Some businesses receive calls where an emergency is more likely. If you operate in one of these verticals, you have additional responsibilities under our Permitted use section:
We may decline to onboard customers in these verticals if we can’t confirm a separate-line-for-emergencies plan is in place, and we reserve the right to suspend accounts that don’t comply.
When the agent recognizes that a caller is indicating an emergency on a call to your business:
/app/calls shows a banner: “An emergency was detected on a call. The agent directed the caller to 911. Review the call →”Review the call and follow up if appropriate — for example, calling the caller back to make sure they reached help. But again: detection is best-effort, not guaranteed. The agent operates on imperfect signals and may not recognize every emergency phrasing, every language, every accent, every call-quality condition. The redirect line is the floor of what the agent does when emergency is recognized; it is not a guarantee that an emergency will be recognized at all. If a caller dies or is seriously harmed because they tried to use the agent instead of dialing 911, the responsibility is not transferred to us by virtue of you having reviewed the dashboard banner. The disclosures in this policy, the contractual prohibitions in the Permitted use section, and your acknowledgments are what govern.
Some context for why this page is structured the way it is.
Federal laws — Kari’s Law (2018) and RAY BAUM’s Act (2018) — set rules for multi-line telephone systems (MLTS) about direct 911 dialing and dispatchable location information. Call My Agent is not an MLTS. The carrier that delivers a call to our system (typically the customer’s own carrier or Telnyx) handles 911 capability at the carrier level. We sit on top of that as an application; we are not in the 911 path.
But callers may try to reach 911 through us by accident — by dialing the business number out of habit when they should have dialed 911, by panicking, or because they don’t know any other number. Our job in that scenario is to redirect them to 911 fast and unambiguously, not to attempt a response.
Wrongful-death and negligence theories in voice-AI cases have not been deeply tested in 2025-2026. Our posture — explicit non-911 disclosures, hard-coded emergency redirect script, account-holder acknowledgments captured at sign-up, dashboard transparency — is built around the principle that the user must always know we are not the emergency path, and that knowledge must be reinforced at every touchpoint.
If a caller experienced an emergency on a call to your business and you believe our system failed to handle it correctly (failed to detect, said the wrong thing, didn’t notify you):
We investigate every report. We may need to access the transcript and recording to do so (per the Privacy Policy). If we find a defect in emergency-intent detection, we fix it.