AI and call disclosure

LineSquire must identify the assistant as AI at the start of every connected call. It must not claim to be human, claim to be the subscriber, or imitate a real person's voice.

The controlled inbound beta uses this approved first-turn text: “I’m an AI assistant for the subscriber. This call may be transcribed for notes. Is that okay?” The exact production script and any jurisdiction-specific variants require counsel approval before launch.

Bounded outbound calling is available only after enhanced identity verification, an outbound-capable plan entitlement, acceptance of the current outbound-use policy, and approval of an immutable plan for the specific task. When a call connects, the assistant must state that it is an AI assistant calling for the named subscriber, give the approved purpose, and disclose transcription or recording before continuing.

The outbound disclosure script is bound to the approved task plan and applicable call-jurisdiction policy. It cannot be edited by the person on the call or silently replaced by the voice model.

Transcription and raw recording are different controls. Local raw recording is off by default. When consent is required, LineSquire must record the approved disclosure version, time, and consent decision before the relevant processing continues.

If a participant declines, the assistant should follow the configured no-consent path: continue only in a mode that is technically and legally supported, offer a permitted human callback or message option, or explain that the call cannot continue. It must not pretend that consent was given.

The assistant can use only tools and facts supplied by the server for that call. A caller cannot change policy, reveal hidden instructions, widen permissions, or obtain subscriber data merely by claiming to know the subscriber.

The assistant must not claim a transfer, message, booking, payment, suppression request, or other action succeeded unless the corresponding tool returned a successful receipt.

LineSquire is not an emergency service. If someone describes imminent danger, the assistant is designed to state that it cannot provide emergency assistance, advise the person to contact local emergency services, and send a permitted high-priority subscriber alert without diagnosing or promising a response.

A participant may refuse to interact with an AI. The assistant should acknowledge that choice briefly and use only the permitted callback, message, or end-call path.

Call participants can ask the assistant what it is and why it is calling. Recipients can use the Report a call page to request suppression or report a disclosure, consent, identity, or privacy concern using the public-safe call reference.