AI Privacy Considerations

Monte Carlo's AI features follow all standard Monte Carlo security practices outlined in Infrastructure Security, Application Security, and AI Security & Governance. This page describes AI-specific data handling, customer controls, and privacy considerations for features powered by large language models.

AI-Specific Data Handling

AI features don't require new data collection or different warehouse access—they work with the same data (information) Monte Carlo already collects for core observability.

For details on what data is used and how it flows through AI features, see AI Architecture & Data Handling.

Privacy Safeguards

AI features may analyze sample data that could contain sensitive information, depending on what customers choose to monitor. Monte Carlo implements these safeguards:

• Stateless processing: AWS Bedrock processes requests without persistent storage; samples discarded immediately after inference (typically seconds)
• Trace redaction: Samples excluded from observability logs when sent to LLM
• No model training: AWS Bedrock contractually commits never to use customer data for training and Monte Carlo does not train or fine tune LLMs.
• Conversation memory: Context for conversational features retained up to 30 days, then purged

Customer Controls

Data Sampling Toggle

Customers can enable or disable data sampling. When disabled, AI features that require samples become unavailable, while metadata-only features continue operating. This provides granular control over what data AI features can access.

AI Feature Opt-Out

Customers can disable specific AI capabilities. Disabling AI features does not affect core observability functionality.

Warehouse-Level Permissions

Standard database permissions control what data Monte Carlo can access. AI features work only within the data access boundaries customers define through warehouse configuration.

Data Residency

Data is processed in the region where your Monte Carlo instance (not data store) is deployed.

Related Documentation:

• AI Security & Governance - Security controls and governance framework
• LLM Training & Observability - Model usage and monitoring
• AI Architecture & Data Handling - How data flows through AI features
• AI Technical Specifications - Model details and system requirements