What does the LLM actually see from our data?
By default the model sees your catalog: table names, column names, types, and a few example values per column. It never receives your tables, and it only sees query results if a user turns on AI analysis.
The catalog, not the tables. Here is exactly what that means.
Sent to the model by default
Table names, column names, data types, the relationships between them, and a few example values per column (for instance the distinct values of a status column, or one sample of a date format). SQL generation needs those examples to filter on the right value and parse the right format.
Not sent to the model by default
Your tables, and the rows your query returns. Results pass through dataface.ai to the screen of the person who asked, and the model does not see them.
The deliberate exception
AI analysis of a result is opt-in. When a user turns it on, the model receives a small sample of that result (up to ten rows) to write its commentary. It is a visible switch, not a default and not a setting buried in an admin panel.
Why this survives a security review
Reviewers ask what is the worst thing that leaves the building. Here it is the catalog and a handful of example values, never a copy of your production tables. Where even example values are too much, catalog sampling can be scoped with the security team during the pilot, and dataface.ai Edge keeps execution on your own infrastructure.
Where this comes from
See it against your own data
A pilot is scoped to one governed use case and time-boxed to eight to twelve weeks, with success criteria agreed before it starts.
Get started freeRelated questions
- What is dataface.ai?dataface.ai is on-source conversational analytics. Ask a question in plain English, and it generates read-only SQL from your schema metadata and runs it on the systems you already own. No warehouse copy, no pipeline.
- How do you run analytics without moving our data?dataface.ai reads your schema and a few example values per column, generates read-only SQL, and executes it on the source system itself. Your tables are never copied: no pipeline, no warehouse copy, no second place to secure.
- How do you stop an AI from writing to or damaging our database?Generated SQL is SELECT-only. INSERT, UPDATE, DELETE and DDL are blocked, queries are parameterized against injection, and every request passes both a workspace gate and a datasource gate.
- Can we run this entirely inside our own network?Yes. Enterprise agreements cover dedicated, VPC, and on-premise deployment, with edge execution on your own infrastructure at a reduced credit rate.