Confidentialité et conservation des données
Alteryx conserve un historique des messages pendant une durée de 90 jours. Pour plus d'informations sur la sécurité et la confidentialité des données avec Alteryx, consultez les Questions fréquentes pour Ask Alteryx ou consultez la page Alteryx Trust.
If you use Ask Alteryx in the Alteryx One desktop app on Windows, local workflows and their outputs can stay on your computer after you delete a chat. For more information, go to Manage Ask Alteryx Chats.
Comment Alteryx Copilot génère des réponses avec Google Gemini
Ask Alteryx génère des réponses à l'aide des modèles Google Gemini hébergés sur la plateforme Vertex AI de Google Cloud. Par conséquent, les contrôles de sécurité, de confidentialité et de traitement des données de Copilot sont basés sur ceux proposés par Google Cloud.
Comment Alteryx protège vos données avec les services Google Cloud
Alteryx configure les services Google Cloud de sorte que les données Copilot soient chiffrées en transit et au repos à l'aide des protocoles de chiffrement standard du secteur. Pour obtenir des informations exactes et à jour sur les pratiques de sécurité et de protection des données de Google Cloud, veuillez consulter :
Sécurité Google Cloud : https://cloud.google.com/security
Gouvernance et confidentialité des données sur Vertex AI : https://cloud.google.com/vertex-ai/docs/general/data-governance
Présentation du chiffrement Google Cloud : https://cloud.google.com/security/encryption
Détails supplémentaires
Par défaut, Ask Alteryx n'a accès qu'aux métadonnées de workflow, au XML de workflow et aux métadonnées de votre jeu de données. Les métadonnées sont uniquement envoyées à Ask Alteryx lorsque vous envoyez une nouvelle instruction. Si vous activez la conscience des données, Ask Alteryx aura également accès aux données de votre workflow.
Information About Your Assets
By default, Ask Alteryx has access to descriptive information about the workflows, datasets, and connections in your workspace. Ask Alteryx uses this information to understand your workflow and to find existing assets that fit your request. It doesn't read the rows of data stored in your datasets. This information includes:
Workflow metadata and workflow XML, including tool configurations and the connections between tools.
Dataset metadata, such as dataset names, source names (for example, a table or file name), column names and data types, and dataset and column descriptions. Descriptions can be AI-generated.
Workflow names, descriptions, and AI-generated workflow summaries.
Summary information about dataset contents, such as column profiles. This helps Ask Alteryx rank datasets by how well they fit your request.
Connection metadata for supported connections, such as connection names and the data objects available through them.
Ask Alteryx only returns assets you have permission to access. Ask Alteryx sends this information to the model only when you send a new prompt.
Data in Your Workflows (Data Awareness)
If you turn on Data Awareness, Ask Alteryx can also sample data from your Designer and Live Query workflows. Ask Alteryx samples data only at the tool anchors it needs to answer your question.
Designer: Ask Alteryx reads the temporary Browse Everywhere snapshot, limited to 1 MB per anchor by default.
Live Query: Ask Alteryx runs supported tools against your connected cloud data warehouse and samples up to 500 rows.
To learn how to turn Data Awareness on or off, go to Conscience des données.
Model Training and Prompt Engineering
Ask Alteryx doesn't train, retrain, or fine-tune AI models with your workflows or data. These points clarify how Ask Alteryx uses AI models:
What Ask Alteryx Does Not Do
Train or retrain foundation models (large language models).
Fine-tune models using your data.
Build new models from scratch.
What Ask Alteryx Does Do
Use pre-trained third-party models.
Improve responses through prompt design and system instructions.
Supply structured context with each request.
Retrieve relevant product or workflow information at runtime to ground responses.
Evaluate and refine prompts and orchestration logic to improve accuracy, safety, and usefulness.
Why This Matters
Model training and fine-tuning require large datasets and permanently change model behavior. Prompt engineering and retrieval don't modify the model. They shape individual requests and responses.
From a compliance perspective, this distinction is critical. Ask Alteryx behavior is driven by runtime instructions and context, not by retraining models on your data.