Skip to content
Operon AI
Legal

AI Disclosure

Last updated July 1, 2026

This page describes how Operon agents identify themselves to the people they speak with, and the commitments we make about that behavior. It applies to every deployment, on every plan, in every jurisdiction.

The commitments

Four things are true of every Operon agent, and none of them are configurable off.

  • It identifies itself as an AI whenever asked, in plain language, without deflecting
  • It discloses upfront wherever a jurisdiction requires it, defaulting to the strictest applicable standard
  • It transfers to a human when someone asks for one, without a retention script
  • It never claims to be a specific named human being

How jurisdiction is determined

Disclosure rules are applied based on the caller's number, the business location handling the interaction, and, where those conflict, whichever standard is stricter. Recording consent follows the same logic, which means a call between a two-party consent state and a one-party state is treated as two-party.

We maintain the ruleset and update it as legislation changes. Customers do not have to track this themselves.

Voice cloning

We support cloning a specific person's voice only with that person's documented, revocable consent, and only for use by the business they work for. We do not clone voices from public recordings, and we do not permit a cloned voice to be presented as a different person than the one who consented.

What agents will not do

Certain behaviors are blocked at the platform level rather than left to configuration or prompt discipline.

  • Deny being an AI when directly asked
  • Give medical, legal, or financial advice
  • Make guarantees about treatment outcomes, case results, or investment returns
  • Use urgency or scarcity claims that are not true of the business's actual availability
  • Continue outreach after any opt-out signal, in any phrasing we can recognize

Human escalation

Every deployment must configure a human escalation path before it can go live. There is no configuration in which a caller who asks for a person receives no path to one. If nobody is available, the agent commits to a callback with a specific window and creates the task.

Reporting a concern

If you interacted with an Operon agent and something about it concerned you, write to us at trust@operon.ai. We investigate every report, we tell you what we found, and where the behavior was wrong we fix it across every deployment rather than only the one you encountered.