DATPROF Analyze provides optional AI profiling functionality that uses a user-configured Large Language Model (LLM) connection to identify the type of data contained in database columns.
AI profiling is optional. DATPROF Analyze does not connect to an external AI service unless an LLM connection has been configured and the user explicitly selects AI profiling. Users can continue to use Local profiling without an LLM connection.
DATPROF Analyze supports connections to Azure Foundry, OpenAI, Anthropic, Google Gemini, and Ollama. The configured endpoint, model, credentials, and provider determine where the AI profiling request is processed.
Data sent during AI profiling
When AI profiling is used, DATPROF Analyze sends information about the selected columns to the configured LLM. This includes:
-
Column information: Information required to identify and profile the column.
-
Sample values: Up to the 20 most common values found in the column.
-
Profiling instructions: Instructions required for the LLM to determine the type of data represented by those values.
Because actual column values are included, data sent for AI profiling may contain personal, confidential, sensitive, or otherwise protected information, depending on the contents of the database.
Data processing and privacy
DATPROF does not control how a configured third-party LLM provider processes, stores, or retains data submitted to its service. Data handling depends on the selected provider, account type, model, region, configuration, and contractual arrangements.
For example:
-
Azure Foundry: Microsoft states that prompts and completions for models sold by Azure are not made available to the underlying model provider and are not used to train generative AI foundation models without permission or instruction. [learn.microsoft.com]
-
OpenAI: API data is not used to train OpenAI models by default. OpenAI documents that abuse-monitoring logs may be retained for up to 30 days by default, with different controls available to eligible customers. [developers...openai.com], [openai.com]
-
Anthropic: Anthropic states that retained API data is not used for model training without express permission. Retention depends on the API feature and model being used. [platform.claude.com], [support.claude.com]
-
Google Gemini: Data handling depends particularly on whether paid services and specific API features are used. Google states that prompts and responses submitted through paid Gemini API services are not used to improve its products, while retention may still occur for purposes such as abuse monitoring. [ai.google.dev], [ai.google.dev]
-
Ollama: Ollama can expose an OpenAI-compatible API for models running locally. When using a local Ollama instance, requests can therefore be processed by that local environment instead of being sent to one of the external providers listed above. [ollama.com]
User responsibility
Before enabling AI profiling, users and organizations are responsible for determining whether the selected LLM provider and configuration are appropriate for the data being profiled. This includes reviewing the provider's current privacy policy, data-processing terms, retention policies, data residency requirements, and any applicable organizational or regulatory requirements.
AI-generated profiling results should be treated as suggestions. Results may be incorrect or incomplete and should be reviewed before being relied upon.