AI Chat and Access Control
When a user opens the chatbot, WordPress sends the Mobie AI Bot API the information needed to check what content that user is allowed to access. This may include information such as the selected avatar, the logged-in user identifier and the course or magazine materials connected to that user.
Messages sent in the chat are processed by the Mobie AI Bot API in Microsoft Azure in the Sweden Central region. To understand questions and generate answers based on the materials available to the user, the chatbot uses large language models and embedding models hosted on Microsoft Azure AI Foundry. The specific models may change over time as the service is improved.
Prompts and responses processed through Microsoft Azure AI Foundry are not used to train or improve the underlying models, and they are not shared with the model providers.
Answers are generated by AI and may contain errors or inaccuracies. They should not be relied upon as the sole basis for decisions, and important information should always be verified from the original materials.
Course and Magazine Materials
Course attachment files, such as PDF, XLSX and PPTX files, are transferred to Azure Blob Storage in the Sweden Central region for processing. These source files are stored there temporarily and deleted after processing.
Course content and zine edition content are sent directly from WordPress to the Mobie AI Bot API for processing.
The material is processed into a form that allows the chatbot to search it and use it when answering questions. The processed data remains stored in the Mobie AI Bot API database until the owner of the material or an administrator removes the related material from the WordPress admin page, after which the data is permanently deleted.
Voice and Avatar Features
If voice input, speech output or avatar features are used, the related audio, speech and session data are processed by the Mobie AI Bot API and Microsoft Azure services in the Sweden Central region so that those features can be provided.
Audio and session data used for speech and avatar functionality are processed only to provide the requested feature and are not kept longer than necessary for the session, unless a longer retention period is required for service security, error investigation or legal obligations.