Chatbot governance — prove your customer-service AI disclosed, and was authorised to act
A customer-service chatbot is not just answering — it discloses it is an AI, issues refunds, makes promises, and escalates complaints. Each is a consequential action a regulator or a court can later question. KYE Protocol™ sits at that moment and proves the chatbot told the customer it was an AI, and that every refund, promise, or escalation was authorised — turning your word against the customer's into a signed, replay-provable record. It cuts a disputed-interaction investigation from days to minutes.
What KYE™ proves on every chatbot turn
Your customer-service chatbot acts inside a customer relationship, so KYE™ governs the turns that carry consequence — not the small talk.
- AI disclosure happened. The "you are talking to an AI" notice required by EU AI Act™ Article 50 is emitted and recorded for the session — you can prove the customer was told, not just claim it.
- Refunds and goodwill were in authority. A refund or credit passes Action Admissibility™ against a per-agent cap and policy version before it is issued — over-cap or out-of-policy is refused at the boundary, not clawed back later.
- Promises were bounded. A commitment the bot makes to a customer is scoped by Purpose Permission™; an ungrounded promise is blocked before it is sent.
- Escalations and account actions are evidenced. Every escalation, account change, or data-access turn emits an Evidence Pack™ — purpose, actor, scope, decision, outcome — verifiable from public keys alone.
- The record is final. Authority Finality™ means a disputed turn can be reconstructed exactly as it happened and contested — your defensible answer to an ombudsman or court.
The disclosure laws your chatbot already has to satisfy
If you run a customer-service chatbot in regulated markets, AI-disclosure is no longer optional — and a screenshot is not evidence. KYE™ maps each obligation to a recorded control on the live session.
| Obligation | What it requires | What KYE™ records |
|---|---|---|
| EU AI Act™ Article 50 | People must be told they are interacting with an AI system, clearly and in good time. | A per-session disclosure event with timestamp, locale, and the wording shown. |
| California SB 1001 (Bot Disclosure Act) | A bot must disclose it is not human when influencing a commercial transaction. | The disclosure bound to the transaction turn it preceded. |
| Colorado AI Act (SB 24-205) | Consumer notice and records for AI systems in consequential decisions. | An API-grade evidence trail per consequential turn. |
| Utah AI Policy Act | Clear disclosure of generative-AI use in regulated consumer interactions. | The disclosure state and the policy version that required it. |
KYE™ does not write your chatbot or replace your CX platform. It governs the authority and evidence of the consequential turns, and ships the proof. The conversation stays yours.
See it, then ship it
You can run the whole gate in your browser before you talk to us — no signup, synthetic traffic only.
- Live sandbox. The chatbot-authority demo runs the disclosure and action-admissibility checks on a sample customer-service exchange and seals a real Evidence Pack™.
- Above any chatbot runtime. Chatbot governance is one solution on the KYE Agent Authority Stack™ — bring your own bot framework; KYE™ adds the authority, evidence, and finality layer it leaves out.
- Pilot in weeks. A focused pilot wires KYE™ to one customer-service flow and produces regulator-ready evidence from day one.