XENI AI Terms

Change log

1. What XENI is and what it is not

1.1. XENI is an assistant. It processes the User's instructions and the context of the active company to generate answers, drafts and proposed actions.

1.2. XENI is not a decision maker. It does not make decisions with legal or significant effects on natural persons, does not evaluate or score people, does not perform biometric identification and does not replace the User's professional judgement in accounting, tax, employment or legal matters.

1.3. Every action that modifies data (issuing a voucher, recording a payment, changing a master record) requires the User's explicit confirmation before it is executed. XENI can prepare; the User decides.

2. Content and ownership

2.1. The inputs the User submits to XENI and the outputs XENI generates are Customer Content for all purposes of the Terms of Service. The Customer owns both and may use them freely.

2.2. Where the Customer has agreed with a third party that ownership of the outputs generated in the course of a service belongs to that third party, XEN respects it: Xeneural claims no right over XENI outputs.

3. No training, with a declared provider chain

3.1. Xeneural will not use Customer Content to develop, train, retrain, fine-tune or distil any artificial intelligence model, its own or a third party's. Xeneural will not allow its subprocessors to use Customer Content for those purposes.

3.2. Xeneural does not train or fine-tune its own models. XENI runs on Google's Gemini models, listed as a subprocessor, contracted on paid terms from Xeneural's cloud project, under terms that prohibit Google from using prompts and responses to train or improve its models. Any change of provider is published on the subprocessor list with thirty (30) days' notice.

3.3. Retention at the provider. The context sent to the model provider is used to generate the answer. Xeneural does not enable memory or extended storage features at the provider. The provider's systems may retain prompts and responses for a limited period, solely to detect uses prohibited by its policies.

4. What context XENI sees

4.1. XENI accesses only the data of the active company in the session and with the same permissions by module, action and site as the User who asks. It cannot see data of the Customer's other companies or of other customers.

4.2. The context sent to the model provider is the minimum needed to answer the query: XENI does not send the whole database, but the records relevant to the question.

4.3. The company's administrator decides whether XENI is active for their company and, per user, what it can query or propose.

4.4. AI-free mode. The company's administrator can switch XENI off for the entire company at any time, with immediate effect. With XENI switched off, no query, suggestion or document of the company is sent to a model provider, and Users do not see the assistant. The administrator can switch it back on whenever they wish. Every change is recorded in the activity log.

5. Transparency and labelling

5.1. The User knows at all times that they are interacting with an artificial intelligence assistant: XENI identifies itself as such in the interface.

5.2. All content that XENI generates and that can leave the platform (documents, reports, exportable drafts) carries a visible indication that it was generated with artificial intelligence assistance and must be reviewed by a person before use.

5.3. XENI explains, at the User's request, which data an answer or a proposed action was based on, in understandable language.

6. Limitations and User responsibility

6.1. Language models can produce incorrect, incomplete or outdated information. XENI presents its answers as support, not as verified truth. The User is responsible for reviewing any output before using it, especially in accounting, tax, employment, legal or security matters.

6.2. Xeneural applies measures to reduce bias and errors (limited context, human confirmation of actions, verification of generated documents) and improves those measures continuously, but does not guarantee that outputs are error-free.

6.3. It is prohibited to use XENI for the purposes forbidden by the Acceptable Use Policy, and in particular to generate unlawful content, to try to extract other customers' data or to circumvent access controls.

6.4. Safety filter. Before sending a query to the model provider and before showing a response, XENI applies an automatic filter that removes payment card numbers, keys and passwords; neutralises instructions hidden in the data it consults; withholds results containing another company's data; does not send to the provider queries that the Acceptable Use Policy prohibits; and withholds responses with harmful content. Every action of the filter is recorded in the activity log, without the affected content. The filter reduces risk; it does not replace the User's review (6.1).

7. Logging and evidence

7.1. Xeneural logs every XENI execution (who, when, on which company, which module, which action it proposed and whether the User confirmed it) with the same retention as the platform's activity log. That log is available to the Customer's administrator and serves as evidence in audits.

7.2. Xeneural uses XENI Service Data (usage counts, errors and response times), aggregated and without identifying individuals, only to operate and improve the Service.

7.3. At the Customer's request, Xeneural delivers the documentation that demonstrates compliance with these Terms: model providers, applicable terms, security measures and, under confidentiality, the execution logs of its account.

7.4. Retention of conversations. Conversations with XENI are kept for as long as the User does not delete them. When a person leaves the company, their conversations in that company are locked and deleted together with their messages after thirty (30) days, unless the administrator places them under legal hold or purges them earlier.

8. Regulatory compliance

8.1. XENI is designed and operated in accordance with the principles of Peru's Law 31814 (ethical, transparent, responsible and privacy-respecting use) and its Regulations. Xeneural classifies XENI as a non-high-risk system: it does not take part in the high-risk uses the law lists (recruitment, credit scoring, health, education, justice, critical infrastructure). If a Customer intended to use XENI for one of those purposes, it must consult Xeneural beforehand and XENI may refuse to execute the action.

8.2. For customers subject to the European Union Artificial Intelligence Regulation, XENI is not used in prohibited or high-risk systems within the meaning of that Regulation.

8.3. For customers that require prior approval of the use of artificial intelligence on their data, Xeneural facilitates that approval by providing the information in this document, the subprocessor list and the Security Summary, and delivers that customer's company with XENI switched off (AI-free mode, 4.4) until it receives it.

9. Changes

Xeneural may update these Terms with thirty (30) days' notice. Any change that extends the use of Customer Content by XENI or its providers requires the Customer's express consent and is never applied by silence.

In two sentences, for those who do not want to read it all

Your data is yours and XENI does not train on it: neither Xeneural nor its model provider. It is in the contract, not just on this page. XENI only sees the active company with the User's permissions, identifies itself as an AI, labels what it generates, executes nothing without the User's confirmation, and the administrator can switch it off for the entire company.