> For the complete documentation index, see [llms.txt](https://docs.userlane.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.userlane.com/security-and-privacy/ai-features-what-data-is-processed-and-where.md). # AI features: what data is processed and where What data the Assistant sends, which model processes it, where it runs, and what is never stored. Userlane uses artificial intelligence (AI) to help your users find answers quickly and to help your teams create content faster. This page explains what the AI features do, what data they process, where that processing happens, and the controls you have as a customer. It covers the AI Assistant, the AI content features, and the data protection principles that apply to all of them. ## The AI Assistant The AI Assistant is a conversational feature within Userlane Assistance. A user asks a question in their own words, and the Assistant replies and, where relevant, links to the guides, announcements, or documents that help them complete the task. It is a discovery layer on top of your existing Userlane content, not a replacement for it. A successful interaction usually ends with the user starting a guide or opening a document. ## Enablement and scope The Assistant is off by default. An administrator turns it on in the portal[^1] using the **Agentic Assistance** setting and applies it to a chosen audience[^2], so you decide which users see it. You can start with a small test group and widen access later. ## AI content features Userlane's Content Agent helps managers produce content faster. It can generate guide text from scratch in your chosen language, refine text you have already written to make it more concise, clearer, friendlier, or more professional, and correct spelling and grammar as part of that refinement. Combined with automatic translation, it can produce localized versions of your content in several languages. These features operate on your Userlane content, not on end user personal data, and generated or refined text is always yours to review before you apply it. ## What data the AI features process For the Assistant, the data processed is the text of the question the user submits and the limited session context needed to keep a conversation coherent. For end users this can include a pseudonymous user identifier and session data. The Assistant does not need a user's name or email address to work. For the content features, the data processed is your Userlane content, such as guide text, and not end user personal data. ## Where processing happens and who is involved All AI processing runs inside Userlane's private Microsoft Azure tenant. Model inference uses Microsoft's enterprise Azure AI service, which is a private deployment of the model that runs inside Userlane's own environment. It is not the public OpenAI service, and no data is shared with OpenAI. Data is hosted in the region tied to your data residency choice. The subprocessor involved in the AI features is Microsoft, which provides the Azure hosting and the model inference environment. The complete list of Userlane subprocessors is maintained in the subprocessor list. If your organization [connects its own AI service](/assistance/the-assistant/connections/connect-your-own-ai-service.md) to the Assistant, the user's question is also sent to that service, which your organization operates. Each user is verified through their Microsoft session, so the service returns only content that user is allowed to see. ## What is stored, and what is not The Assistant conversation is not kept in the user's browser after the session ends. Conversation content is, however, processed and stored within Userlane's private Azure environment so that the feature can work and so that administrators can review it. Administrators can see Assistant conversations in **AI Assistant Analytics** in the portal. Chats are grouped into topics so you can see what users ask about, and individual conversations can be opened and read. This helps you find content gaps and improve your guidance. Operational logs are retained under Userlane's standard retention policy. Userlane does not use your data, or your users' data, to train AI models. ## Your controls You can enable or disable the Assistant and scope it to a specific audience. You can choose your data residency region. You can review Assistant conversations in analytics. Userlane's analytics privacy settings also apply to the data collected across the product. ## Data protection principles These principles apply to all Userlane AI features. **Your data is your data.** Userlane processes it to provide the service, following your instructions as the data controller. **Secure in transit and at rest.** Userlane applies encryption, access controls, monitoring, and the other measures described in [Hosting and certifications](/security-and-privacy/hosting-and-certifications.md). **No sale, and no sharing for others' purposes.** Userlane does not sell your data and does not share it with third parties for their own purposes. To run the service, Userlane relies on a small number of subprocessors that process data on Userlane's behalf under contract. ## How answers are grounded The Assistant answers from your own content: guides, announcements, and documents, any [knowledge bases you connect](/assistance/the-assistant/connections/knowledge-connectors.md), and your own AI service if you connect one. When a question matches that content, it replies and links to the relevant item. When nothing matches, it can answer from general knowledge, labeled as such, or point the user to your support contact. General knowledge is on by default and can be switched off per application. The Assistant does not browse the open web. ## Human oversight and accuracy AI-generated answers can be incomplete or wrong, so your content remains the authoritative source. Administrators keep oversight through **AI Assistant Analytics**, where conversations can be reviewed and content gaps corrected. Content Agent output is always reviewed and approved by a manager before it is published. ## Guardrails and safety Guardrails run server-side and cannot be bypassed from the browser. They include content filtering on inputs and outputs, rate limiting per user and per tenant, and handling for prompt injection and misuse attempts. The Assistant operates only within the scope an administrator enables and cannot take actions in your systems. ## Model and provider details Userlane does not disclose specific model versions, which change as models are updated. Any change to the provider or platform is managed under Userlane's change process and reflected in the subprocessor list. ## Supported languages The Assistant is configured for English and German, which are the languages it is tested and tuned for. Users can ask questions in other languages and the Assistant will respond, but answer quality is not guaranteed outside English and German. Content Agent generation and translation work across a wider set of languages. ## Data subject rights Under the General Data Protection Regulation (GDPR), the customer is the data controller for the personal data processed through Userlane, and Userlane acts as the data processor. This shapes how data subject requests are handled. End users who want to exercise their rights, such as access, correction, or deletion, should contact their own organization, which is the controller of their data. The customer decides how to respond. Where a request covers data processed by the AI features, Userlane supports the customer in fulfilling it as the processor. This includes locating and, on the customer's instruction, deleting Assistant conversation data held in Userlane's environment, including the conversation content visible in **AI Assistant Analytics**. Customers can raise a data subject request that involves Userlane by contacting . Requests are handled in line with the [data processing agreement](https://userlane.com/dpa) and applicable law. For more detail on how Userlane processes personal data, see the privacy policy and the data processing agreement. ## EU AI Act position Userlane assesses the Assistant as a limited-risk AI system under the European Union Artificial Intelligence Act (EU AI Act), subject to transparency obligations rather than high-risk requirements. Users are informed they are interacting with an AI system. Userlane monitors implementation of the Act and will update this page as obligations take effect. ## Compliance Userlane is certified to ISO/IEC 27001. Userlane also monitors and prepares for other AI-specific regulation. ## Relevant articles {% content-ref url="/pages/XDGntxUJimFOtJOihyow" %} [The Assistant](/assistance/the-assistant.md) {% endcontent-ref %} {% content-ref url="/pages/jaGpQWjETS3dwv2yneLk" %} [Configure the AI Assistant](/assistance/the-assistant/configure-the-ai-assistant.md) {% endcontent-ref %} {% content-ref url="/pages/7EPCKtnfdYXjGlW43dbR" %} [Connect your own AI service](/assistance/the-assistant/connections/connect-your-own-ai-service.md) {% endcontent-ref %} {% content-ref url="/pages/PeD6svglTFegZfmDCLJA" %} [Data flow diagram](/security-and-privacy/data-flow-diagram.md) {% endcontent-ref %} {% content-ref url="/pages/h4XuR33OT54sP0N63SOo" %} [Hosting and certifications](/security-and-privacy/hosting-and-certifications.md) {% endcontent-ref %} [^1]: The portal is the web application where administrators configure Userlane, target content, and view analytics. [^2]: An audience is a group of users defined by attribute rules, used to decide who sees Userlane content. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation by asking a question. Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters: ``` GET https://docs.userlane.com/security-and-privacy/ai-features-what-data-is-processed-and-where.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.