Texting customers is regulated (TCPA and carrier rules), and Sovereign's SMS layer is built fail-closed: if a compliance check can't be verified, the message does not send.
STOP always wins
When any recipient replies with STOP, UNSUBSCRIBE, OPTOUT, or OPT OUT as a standalone word:
- The keyword is handled before any AI reply is generated — the assistant never talks past an opt-out.
- The number goes on the suppression list immediately.
- A single opt-out confirmation is sent, and nothing else.
Replying START or UNSTOP reverses it and records renewed consent. You don't configure any of this — it's always on and applies even to numbers texting in for the first time.
What every outbound text passes through
Before any campaign or automated SMS goes out, each recipient is checked against:
- The suppression list — STOP replies and bounces land here; suppressed numbers are filtered out. If the suppression lookup itself fails, the send is blocked rather than risked.
- Consent — recipients need a recorded consent basis for the message type.
- Quiet hours — messages are held outside acceptable local sending hours.
- Cooldowns and cost guards — rate and budget limits prevent runaway sending.
Disclosure applies to texts too
When the AI assistant is the one texting (for example, a chat conversation continuing over SMS), the AI identifies itself — the same disclosure doctrine as voice calls.
What you should still do
- Collect phone numbers with clear language about what you'll send.
- Keep your own promise: if a customer asks a human to stop texting, remove them — don't rely on them texting STOP.
- Never paste purchased contact lists into any send path. Consent you didn't collect is consent you don't have.