> For the complete documentation index, see [llms.txt](https://doc.dastra.eu/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://doc.dastra.eu/en/features/generalites/ai-assistant/ai-assistant-faq.md). # FAQ ### Privacy and data **Is my data used to train AI models?** No. The models used by Dastra are pre-trained and hosted on Microsoft Azure. Your data is never used to train, fine-tune, or improve the base models — by Microsoft, OpenAI, Mistral, or any other third party. This guarantee is contractual. *** **Is my workspace data sent to the model without my knowledge?** No. Dastra only transmits the information strictly necessary for the requested generation: the text you enter, and where applicable the fields of the object you are working on (processing activity, AI system, rights request, etc.). Other records in your workspace, your credentials, and session metadata are never included. For a breakdown by feature, see the How it works page. *** **Where are the AI models used by Dastra hosted?** For the OpenAI, Mistral, and Open source families, models are hosted on Microsoft Azure infrastructure in France. Your data does not transit through OpenAI's or Mistral's own servers. Dastra uses exclusively Azure managed services. If you configure a **Custom AI provider** with your own API key, hosting depends on the provider you choose. *** **Are my prompts retained after generation?** Azure models are stateless: prompts and results are not stored in the model between requests. They are also not logged for commercial reuse. **Exception:** Azure maintains an automated abuse detection mechanism. If content is flagged, a sample may be temporarily retained in a customer-isolated environment for review purposes. For resources deployed in the EEA (which is the case for Dastra), any human reviewers involved are also located in the EEA. This retention remains exceptional. *** **Who can see AI invocations made within my organisation?** Workspace administrators can access the **AI invocation logs** (90-day history) via **Organisation Settings > AI Assistant > AI invocation logs**. These logs show the date, status, operation type, model used, duration, and the user who triggered the generation. The content of prompts and responses is not displayed. *** **Is the AI Assistant covered by Dastra's DPA (Data Processing Agreement)?** Yes. Use of the AI Assistant falls within the existing contractual framework between your organisation and Dastra. Microsoft Azure acts as a sub-processor, bound by Azure's contractual data protection commitments. For more details, refer to your Dastra DPA or contact our team. *** ### Quality and reliability **Can the results generated by AI be trusted?** Generated results are starting proposals, not definitive answers. They accelerate your documentation work but do not replace expert judgement. All generated content must be reviewed and validated by a qualified person before use, particularly for compliance analyses, privacy notices, and responses to data subject rights requests. *** **What is a beta feature?** Some AI features are released in beta: they are operational but continue to be refined. Results may be less consistent than in stable features and should be treated with extra care. Currently in beta: processing activity compliance analysis and AI system compliance analysis. *** **Can AI make mistakes on legal questions?** Yes. The AI Assistant is not a lawyer. Generated content — compliance analyses, notices, risk assessments — is a decision-support aid based on information recorded in your workspace. It does not constitute legal advice and cannot replace the expertise of a DPO or specialist counsel. *** ### Configuration and access **How do I choose the AI model used?** Go to **Organisation Settings > AI Assistant** to select the model family: OpenAI (default), Mistral, Open source, or a custom provider (Custom AI provider). For each family, three model tiers are used automatically depending on the complexity of the task (Fast, Smart, Extended context). *** **Can I connect my own AI provider?** Yes, via the **Custom AI provider** feature. Any provider compatible with the OpenAI API standard can be connected: OpenAI, Google (Gemini), Mistral, Microsoft Foundry, or a self-hosted LLM. Configure your credentials in **Organisation Settings > AI Assistant > Custom AI provider**. {% hint style="warning" %} With a Custom AI provider, the Azure guarantees described in this documentation no longer apply. Data is subject to your provider's policy. {% endhint %} *** **Can I disable the AI Assistant for my organisation?** Yes. Contact our support team to disable the feature at organisation level. Disabling is global: it is not possible to disable only certain prompts while keeping others active. *** **Is the AI Assistant available on all Dastra plans?** Yes, the AI Assistant is available on all Dastra plans. Some advanced features may be subject to specific conditions. Contact our team for any questions about your plan. *** **What are AI credits?** AI credits are the unit of measurement for AI Assistant consumption within your organisation. Each call to an AI feature consumes a certain number of credits, which varies depending on the model tier used (Fast, Smart, or Large context) and the amount of data processed. Administrators can view current consumption and the remaining balance under **Organisation Organisation Settings > AI Assistant** ([direct link](https://app.dastra.eu/general-settings/ai)). *** **What happens if my organisation reaches its credit quota?** When the quota is reached, AI features are temporarily suspended until the next renewal. Your workspace and data remain fully accessible — only AI generation is disabled. To increase your quota or request an early renewal, contact our team. *** **Does the credit system apply when using a Custom AI provider?** No. When your organisation configures a **Custom AI provider**, the Dastra credit system is disabled — no credits are counted. Consumption is managed directly by your provider according to its own billing terms. *** ### AI Act (EU Regulation 2024/1689) **Is Dastra's AI Assistant subject to the AI Act?** Yes. Dastra's AI Assistant falls within the definition of an AI system under the AI Act (EU 2024/1689). Depending on the features used, it is primarily classified as **limited risk**, which implies transparency obligations (Art. 50) when generated content is intended for natural persons. *** **Do I need to inform data subjects that I used AI to respond to their requests?** Not under a strict obligation of the AI Act, but it is strongly recommended as a best practice. Article 50(1) of the AI Act targets AI systems **that interact directly** with natural persons. In Dastra, that is not the case: a human user (DPO, legal counsel) generates the text, reviews it, validates it, and sends it themselves to the data subject. There is no direct interaction between the AI system and the end recipient — the strict obligation of Art. 50(1) does not apply. {% hint style="info" %} The AI Office's draft guidelines (published for stakeholder consultation in 2025) explicitly confirm this interpretation: *"Direct interaction excludes indirect or mediated human interaction where the person is exposed to AI outputs without directly interacting with the system — **for example if customer service representatives use AI assistance tools to help them communicate with natural persons** or if the AI output is not made available to the person by the AI system itself, but by another person disseminating the content."* (Section 3.1.1). The Dastra use case corresponds exactly to this scenario. Note: Article 50 obligations apply from **2 August 2026**. The final guidelines may provide further clarification. {% endhint %} That said, informing the data subject that a response was AI-assisted remains a **best practice** under the GDPR principle of fairness and transparency (Art. 5(1)(a)). Dastra does not automatically add this mention: if your organisation wishes to include it, it is up to you to do so before sending. *** **Is Dastra a provider or deployer under the AI Act?** Both, depending on the use case. When you use the default models (OpenAI, Mistral, Open source via Azure), Dastra is the **provider** of the AI system and your organisation is the **deployer**. When you connect your own provider via Custom AI provider, Dastra acts solely as a compliance tool. Your organisation then assumes the role of provider and/or deployer with respect to its own model. --- # 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 dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter: ``` GET https://doc.dastra.eu/en/features/generalites/ai-assistant/ai-assistant-faq.md?ask= ``` The question should be specific, self-contained, and written in natural language. 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.