Impact Intelligence maintains complete data security through an innovative architecture designed to protect your sensitive information while delivering powerful AI capabilities.
How Your Data Stays Secure
Your actual data values never leave your secure environment. Impact Intelligence uses a metadata-only approach where Impact Assistant receives only:
- Dataset names and their descriptions
- Field names and field types
- Aggregate statistics about fields (not associated with any individual record):
- For date fields: The earliest and latest dates
- For text fields: The 10 most common values
This metadata enables Impact Assistant to understand what questions it can answer and provide meaningful filtering and queries, while ensuring sensitive data remains completely protected.
Example of Privacy Protection
Consider a dataset called "Customer Orders" with thousands of records. If the dataset contains a "Product Type" field where "laptops," "tablets," and "phones" each appear in at least 1% of records, Impact Intelligence will know these are common product types to support queries like "show me laptop sales last quarter."
However, if the same dataset contains a "Shipping Address" field, individual addresses almost certainly won't represent 1% or more of records, so specific addresses are never sent to the AI. This protects personally identifiable information while still enabling powerful data analysis.
Government Cloud (FedRAMP) Security
For Spider Impact government customers, all infrastructure operates within Amazon GovCloud behind our secure FedRAMP-authorized boundary. The AI services are also within Amazon GovCloud and part of this secure boundary. While we are fully authorized to send secure information to these AI services, we maintain our metadata-only approach for additional privacy protection.
Commercial Cloud Security
For commercially hosted customers, we use Amazon Web Services with enterprise-grade security to process metadata. The AI services have contractual agreements that **your data will never be used for AI training**. This guarantee applies to both the metadata we send and ensures your organization's information remains completely private.
Key Security Benefits
- No data values leave your environment - All actual data processing happens securely on your server
- AI cannot hallucinate data - Since all values come directly from your database, Impact Assistant can only show real information
- No AI training on your data - Contractual agreements prevent any use of your information for model training
- High performance - Minimal data transfer means fast, responsive interactions