AI Transparency

Summary

This summary is here to help you. It does not replace the full text below.

• Exubee is the AI assistant inside Stayantra. When you talk to Exubee, you are told that replies are generated by AI, and you can always ask for a person at the hotel instead.
• AI helps with guest chat and voice, suggested replies, hotel operations, marketing images and posts, reading ID documents and receipts, and background research on arriving guests.
• For hotels, AI works in Co-Pilot first: it proposes and a person approves. A hotel can later let a single area run on Auto-Pilot, but only after a proven track record, an explicit opt-in, and inside limits we set.
• AI never issues refunds and never sends WhatsApp messages. People do both.
• Your chat text, booking context and, for ID reading, the image of your document are sent through OpenRouter to companies that run AI models. We allow only companies that keep no copy of the request and do not train on it, and never companies based in China. Details are in section 7 and in our Privacy Policy.
• We do not use AI to recognise faces, identify people from biometric data, or read your emotions.
• AI can be wrong. Check important facts, and never rely on AI for medical, legal or safety decisions.
• You can ask for a human to review any decision that affects you, and you can report harmful AI output to us.

1. About this page

This page explains how Stayantra uses artificial intelligence (AI). It is written to meet the transparency duties in the EU Artificial Intelligence Act (Article 50), the UK GDPR and EU GDPR (including Article 22 on automated decisions), Nepal's Individual Privacy Act 2075 (2018) and E-Commerce Act 2081 (2025), and India's Digital Personal Data Protection Act 2023.

It sits alongside our Privacy Policy, which explains in full how we handle personal data, and our Terms of Service.

2. What Exubee is, and who builds it

Exubee is the AI technology inside Stayantra: the software, methods and designs that orchestrate its AI agents and run its learning loop, memory, context management, autonomy (Co-Pilot and Auto-Pilot) and guardrails, together with its model routing and evaluation tools. You meet it as the assistant you chat with, backed by a network of specialised AI agents that each handle one area of hotel work, such as pricing, marketing, housekeeping, guest messages or events. A coordinating agent routes each request to the right specialist.

Exubee is owned by Nexalaris Tech Pvt. Ltd., the company that developed Stayantra. Nexalaris Tech Pvt. Ltd. has granted the Stayantra companies, [Pending: UK company legal name] and [Pending: Nepal Stayantra company legal name], which provide the service to you, a perpetual and irrevocable licence to use Exubee in Stayantra. If the Stayantra business passes to another company, the licence passes with it, so the Exubee features you use carry on. Your data is not part of Exubee. Policies, memories and decisions that Exubee learns from your data are your data, and Nexalaris Tech Pvt. Ltd. may not use them outside Stayantra. Nexalaris Tech Pvt. Ltd. is a licensor and service provider; it is not the company you contract with. Using Stayantra gives you the right to use Exubee as part of the service, and no other rights in it.

Exubee does not train its own foundation models. It sends instructions and context to large language models run by the providers listed in section 7, and uses their answers.

3. Where you meet AI in Stayantra

4. Our transparency commitments

You are told when you are talking to AI. Every conversation with Exubee is clearly labelled: "Exubee is an AI assistant — replies are generated by AI." We never present Exubee as a human.

Labelling AI-generated images and posts. Stayantra marks images made with our AI image tool as AI-generated in the product, and adds the line "Image created with AI." to the caption when a hotel publishes one to Facebook or Instagram through Stayantra. The hotel is responsible for applying any further label the law or the platform requires (for example Meta's "AI info" label), including on AI media it publishes elsewhere, and our Acceptable Use Policy requires hotels to do so. When the Stayantra companies publish AI-generated images on Stayantra's own channels, we disclose that they are AI-generated.

AI-drafted text sent by a person is that person's message. When hotel staff send a reply that AI drafted, they have read and approved it, and the hotel is responsible for it as its own message.

5. Human oversight

People stay in charge of AI in Stayantra. These controls are built into the product; they are not optional settings a hotel can remove.

• Co-Pilot comes first. Every AI feature starts in Co-Pilot mode: the AI proposes an action and a person at the hotel approves, changes or rejects it. No feature launches in Auto-Pilot mode.
• Auto-Pilot has to be earned. A hotel can move one area of work (a stream, such as pricing or marketing) to Auto-Pilot only after staff have agreed with the AI's proposals at a high rate — 85% to 92% agreement over 20 to 100 decisions in 30 to 90 days, depending on the stream — and the owner has explicitly opted in with a typed signature.
• Limits we set. Even on Auto-Pilot, AI works inside limits that Stayantra sets and hotels cannot loosen, for example a maximum price change of 15% in a day and caps on marketing spend.
• Automatic demotion. If the AI breaks one of those limits, that stream goes straight back to Co-Pilot.
• Kill switch. Each hotel has a switch that stops AI actions for the whole hotel at once.
• Audit trail. Every AI action is written to an audit record that cannot be edited or deleted, and the hotel can see it.
• Things AI never does. AI never issues refunds; a person always decides and sends them (a hotel manager or owner, or, where a hotel fails to refund a Pay with Stayantra booking on time, a person at Stayantra). AI never sends WhatsApp messages. It is also limited to a small, fixed number of steps for any one task, so it cannot run on unchecked.

The hotel remains responsible for its prices, messages and posts, whether a person or the AI prepared them.

6. Limitations: AI can be wrong

AI models produce answers that sound confident but can be incomplete, out of date or simply wrong. In particular:

• Check key booking facts — price, dates, cancellation terms and what is included — on the booking summary shown before you pay. The booking summary, not the chat, is what you agree to.
• Do not rely on AI for medical, legal or safety matters. If you have an allergy or a medical need, tell the hotel directly as well as Exubee. In an emergency, contact local emergency services and hotel staff, not the chat.
• Translations and summaries made by AI may lose meaning. Ask for the original if something matters.
• ID and receipt reading can misread a character. That is why hotel staff check every field.

Nothing on this page limits your rights if AI output causes you loss. Those rights are set out in our Terms of Service, and we never exclude liability that the law does not allow us to exclude.

7. AI providers, countries and your data

All AI model requests are sent through OpenRouter (United States), which passes each one to a hosting provider: a company that runs the model we have chosen for that task. The company that developed a model and the company that runs it are often different.

What this means for your data:

• The personal data in a request is sent to these providers. That can include what you type or say in chat, booking context such as your name, dates and preferences, and, for ID reading, the image of your document.
• We use paid, business access to these services, not free consumer versions.
• With every request, we instruct OpenRouter to use only hosting providers it lists as zero data retention. They do not store the request or the answer and do not use them to train models.
• We instruct OpenRouter never to use a hosting provider based in China. DeepSeek and Kimi are open-weight models that other companies can run, so we use them only as run outside China and your data is not sent to DeepSeek, Moonshot AI or Alibaba.
• If no hosting provider that meets these rules is available, the request moves to our next model or fails. It is never sent to a provider that does not meet them.
• Two exceptions. OpenRouter's speech-to-text service does not accept these instructions, so for voice we use a model whose only hosting providers, DeepInfra and Groq, are both on OpenRouter's zero data retention list. And the public web search used for guest web profiling is run by a search provider through OpenRouter that zero data retention does not cover.
• OpenRouter keeps records of each request, such as the time, the model and the cost. Its policy is not to keep the content of requests unless a customer turns on prompt logging.
• Sending data to countries such as the United States is an international transfer. We protect those transfers as described in our Privacy Policy, which also lists every provider we use.

8. What we do not do

• No face recognition or biometric identification. When AI reads an ID document, it reads only the printed text fields. It does not analyse the photo on the document, compare your face with anything, or identify you from biometric data.
• No biometric categorisation. We do not use AI to infer race, political opinions, religion, sexual orientation or similar characteristics from your data.
• No emotion recognition. We do not use AI to detect or infer your emotions from your voice, face or text.
• No social scoring and no manipulation. We do not use AI to score people's trustworthiness or to push people into decisions through deceptive techniques.

9. Automated decisions and your right to human review

We do not make decisions based solely on automated processing that have legal or similarly significant effects on guests. Bookings, refunds, complaints and account actions against you are decided or checked by a person.

Some things are calculated automatically, for example the order of search results and featured hotels (explained on our ranking page), and, for hotel staff, an average of guest ratings that can make a staff member eligible for a bonus or flag a performance review. In those staff cases the system only makes a recommendation; a human manager always makes the decision.

Wherever AI plays a part in a decision about you, you can:

• ask us or the hotel to explain how the decision was reached;
• give your point of view and contest the decision; and
• ask for a person to review it.

Send the request to info@stayantra.com. If it concerns a decision a hotel made, we will pass it to the hotel and help it respond. You also have the right to object to guest web profiling and to AI memory; if you object, the hotel must stop and we will help it do so.

10. Not a high-risk use of AI

The EU AI Act sets stricter rules for AI used for certain purposes it lists as high-risk. We do not use AI for purposes listed as high-risk, such as deciding access to education or employment, assessing creditworthiness or credit scores, setting life or health insurance prices, deciding access to essential public services, law enforcement, migration or border control, or remote biometric identification.

Staff performance features give recommendations only, and a person decides. If we ever plan a use that falls into a high-risk category, we will meet the requirements that apply before we launch it and update this page.

11. Feedback and reporting harmful output

If Exubee or any AI feature gives you an answer that is wrong, offensive, unsafe or discriminatory, please tell us. Email info@stayantra.com with:

• where it happened (for example marketplace chat, Stayantra Go, or the hotel dashboard);
• the date and approximate time;
• what the AI said or did, with a screenshot if you have one.

We review every report, correct the problem where we can, and use what we learn to improve safeguards. If the output caused you harm or you want a formal answer, use our Complaints process: we acknowledge every complaint and give it a reference number within 1 business day (and never later than 48 hours), and send a written, reasoned decision within 15 days.

12. Changes to this page

We update this page when we add, remove or materially change an AI feature or provider. The date at the top shows the current version. If a change affects how your personal data is used, we will tell you in advance as described in our Privacy Policy.

Questions about this page: email info@stayantra.com.