1. Always labelled as AI
Every message from the AI agent carries an “AI” label. Before its first reply in a conversation, the visitor is told they are chatting with AI. Website owners cannot switch this notice off. The AI agent can have its own name (by default, Cubby), but it never pretends to be a person.
2. Answers from the website’s own information
The AI agent answers from the knowledge base the website owner provides: Q&A pairs, documents and imported web pages. When a question matches a Q&A pair exactly, that answer is sent word for word. Otherwise, the most relevant passages are given to the AI model to write an answer, and the source is shown with the reply.
When there is no relevant information, the AI agent says so and passes the conversation to the website’s team, instead of guessing.
3. A person is always available
Visitors can ask for a person at any time. Website owners can also limit how many answers the AI gives in one conversation, after which it hands over automatically. Outside business hours, visitors can leave a message for the team.
4. What is sent to the AI model
When AI answers are on, the visitor’s current question and the knowledge base passages that matched it are sent to the model provider. When translation is on, the text that needs translating is sent. The full conversation history, the visitor’s context and analytics data are not sent.
Model providers are contractually prohibited from training on this data. AskCub never uses conversations or knowledge bases to train AI models. Website owners may connect their own model provider key instead of the hosted model; their agreement with that provider then applies.
5. What the AI agent does not do
- It does not take actions such as placing orders, issuing refunds or changing bookings.
- It does not give medical, legal or financial advice beyond what the website has published in its knowledge base.
- It does not make decisions about people; it answers questions.
AI answers can still be wrong. Sources are shown so visitors can check, and website owners can correct answers and review the questions the AI couldn’t answer.
6. Responsibilities of website owners
Website owners choose whether to turn on the AI agent, what goes into their knowledge base and how it hands off. They are responsible for keeping that information accurate and for meeting the rules that apply to their business and region.
7. Questions
Visitors with questions about a specific conversation should contact the website they chatted with. For questions about this policy, write to hello@tapcub.com.