AI data sharing

A short guide to what data in Omniscope is shared with AIs. Written regarding Omniscope 2025.1.

You might be concerned about your sensitive company data being shared with another company, or perhaps a foreign government, or perhaps used to train an AI model. We understand and share these concerns, and can assure you that Omniscope will never share data with an AI without your explicit configuration.

Key points

1. AI integration is off by default

Omniscope will not speak to any AIs, unless you have specifically

• Added and configured an AI provider such as OpenAI with the necessary API key
• Further configured an AI integration to use a specific model.

(As documented in How to enable AI in Omniscope).

2. Omniscope only talks to the AI when asked

If the above has been satisfied, Omniscope will only talk to the AI when asked, as follows. It won't talk to the AI in the background without your knowledge.

• Report Ninja:
  The conversation with the AI only happens while the right sidebar ("Report Ninja") is showing. When collapsed, which is the default for all except "Instant Dashboard" projects/reports, the conversation is paused, and no contact with an AI is made.
  
  
• AI Block:
  Talking to the AI only happens when you add an "AI Block" to the workflow, open it, and start chatting. If you don't add that block, or open it, nothing happens.
  
  
• Custom SQL:
  Omniscope talks to the AI only when you add a Database block, configure Custom SQL, enter a question/instruction in "Auto-generate SQL by asking a question", and hit "Convert question to SQL".

3. Omniscope only shares limited metadata 

If the above has been satisfied...

• Report Ninja:
  Omniscope shares only the following metadata about the first data source in the report with the AI:
  • Field name
  • Field data type
  • Field description annotation (free-text, configurable in the Field Organiser block, most commonly empty)
  • Field source annotation (editable in the Field Organiser block, by default containing the filename or database table name of the original data source in the workflow)
    
    
• AI Block:
  Omniscope shares only the field names and data types of the block input data with the AI. Does not share any row-level data (field/cell values, example, sample values).
  
  
• Custom SQL:
  Omniscope shares only the table names and column names/types of the database tables referenced by the Database block configuration, with the AI. Does not share any row-level data (field/cell values, example/sample values).

4. You can lock Omniscope down

The AI settings in Omniscope can be configured in a given folder, and overridden in a subfolder. If you wish to enforce a policy, you should configure your AI providers in the root folder, and untick "Allow subfolder override" to prevent users from using their own providers instead.

You can then limit access who can edit the root folder settings/permissions, to prevent users from changing your centralised policy.

Coming soon: data sharing controls

Future versions of Omniscope will let you set, on a provider basis, the data that can be shared with that provider. 

Note, we expect to start sharing limited cell values (e.g. to give example values for a given field), but only when expressly permitted, with a provider, to improve the experience and allow the AI to have a clearer understanding of the data. Future use cases might even share entire datasets, but we'd always require an explicit opt-in for this.

We intend to provide controls allowing you to configure multiple providers and specify, for example:

• OpenAI - "only share field metadata" (i.e. names/types, as currently / as above)
• DeepSeek - "don't share anything data-related" (e.g. a support assistant)
• Azure OpenAI - "share cell values too"
• A given custom (local) model - "share everything, unrestricted".