Enhancing Privacy in GenAI Applications

Text-to-SQL has LLMs generate SQL queries to retrieve data that can have PII, or quai-identifiers that can be used to infer a person’s identity
Function calling or invoking APIs can also return data with PII in a JSON structure which requires specific forms of privacy enhancement
Document RAG can extract data from vector DBs that contain PII and other sensitive information subject to privacy controls

Ackuity can be configured to apply a wide range of privacy enhancement techniques, including anonymization, hashing, tokenization, encryption, masking, homomorphic encryption and differential privacy. In addition, Ackuity can directly resolve each of the above privacy challenges. Ackuity :

Detects sensitive information subject to privacy controls
Identifies direct PII and quasi-identifiers, and enforces policies that remove or enhance privacy for sensitive information being returned
De-vectorizes extracted data, applies privacy enhancements, and re-vectorizes before sending to an LLM for inference
Allows compliance with frameworks including GDPR, CCPA, PIPEDA, NIST, GLBA, HIPAA, and other regulations and best practices