Collibra AI brings automated features that help you accelerate the discovery, curation, governance, and visualization of your data.
For information on how we leverage AI in our products, visit the Collibra Trust Site.
Where can you find Collibra AI in Collibra?
- Asset recommender
- Create asset description recommendations with Collibra AI
- Data similarity
- DAX analysis for Power BI technical lineage
- SQL assistant for data quality (Beta)
FAQ on generative AI
This FAQ applies only to generative AI features, such as asset description recommendations, DAX analysis, and SQL assistant.
- How does Collibra AI for descriptions work?
The general flow is as follows:
- When a user requests a description of an asset in Collibra, we send the request to a proxy.
- The proxy checks whether the rate limit of a customer instance has been reached and sends the request to the AI model.
- The AI model responds to the proxy.
- The proxy sends the answer to Collibra where the answer is displayed to the user.
- Which data is stored in the AI models?
The only data stored in Vertex is metadata regarding the request, such as total billable characters and character count. Any metadata needed to generate the description is not stored. The user text input and model output are only processed, not stored. - Are the AI models refined or retrained in any way based on requests?
No, the models are not refined or retrained based on the metadata or the model output. - How is data protected when it is sent to an AI model?
All data is encrypted using HTTPS and TLS protocols.