Use of customer data with AI Center

We process datasets, projects, and packages for service provisioning as further described in the About AI Center page from the AI Center user guide. We store data that the customer explicitly uploads, such as datasets for providing the services until deleted by the customer or until the project/tenant/organization is deleted.

We process the stored data as instructed by the customer to retrain ML packages when the customer explicitly starts or schedules the pipeline. Those pipelines also generate data that is visible to the customers. The customer has full control over the data used for training, for example upload or deletion.

On top of artefacts data, training and full pipelines generate a new version of the ML package that is only accessible to the customer and is better adapted to their data.

We process telemetry data, such as usage data, to make sure that our platforms operate smoothly and to optimize the user experience based on the feedback and usage patterns we observe. We use data analytics to measure and evaluate the performance of our services.

We process identity information (from our Identity server) to make sure we have the right permissions in place and that you access only to the projects and pages you should have access to (based on the role-based access control settings set by the customer). We also store and expose audit and logs to the customers at project level.

We may process data that you send to us for support or as necessary for troubleshooting issues, in accordance with our Support terms.

AI Center does not process customer data to train its models. The customer retrains machine learning models and has full control over the usage of training data, but the retrained model is only accessible to this specific customer.