Policy
Secure Use of LLMs
GenDB: Using Enterprise IT Telemetry with GenAI Sensibly and Securely
Last Updated: April 16, 2024
Overview
What is GenDB?
GenDB is NeuBird AI's GenOps-based virtual database that enables secure and sensible read-only access to IT telemetry to solve ITOps problems. It is the foundation of how NeuBird AI uses large language models responsibly, without exposing your sensitive production data or internal intellectual property.
GenDB operates iteratively: it strips sensitive identifiers such as IP addresses and PII before sharing operational context with LLMs for guidance, then independently executes telemetry queries within your secure internal systems. This architecture ensures sensitive data never leaves your VPC while still leveraging AI capabilities for problem-solving.
Two Golden Rules
How NeuBird AI uses LLMs responsibly
Don't send your data to an LLM
The LLM never sees your specific telemetry with sensitive data such as IP addresses and personally identifiable information. Sensitive identifiers are stripped before any context is shared. The platform queries LLMs for advisory guidance while executing all real database queries locally within your VPC.
- Sensitive identifiers (IPs, PII) stripped before LLM context is shared
- Actual queries executed locally within your VPC
- Zero exposure of sensitive operational data to third parties
Don't get IP from an LLM
NeuBird AI does not use LLMs to generate proprietary code for product development and does not extract intellectual property from LLM interactions. We use AI to solve operational problems, not to build competitive advantage on the back of your data or third-party model knowledge.
- No proprietary code generated by LLMs for NeuBird AI products
- No IP extraction from LLM interactions
- AI used for operational guidance, not product development
How It Works
The GenDB workflow
Anonymized scenario sent to LLM
GenDB constructs a hypothetical, anonymized representation of the ITOps problem and sends it to the chosen LLM for advisory guidance.
LLM returns advisory guidance
The LLM responds with suggested query strategies and approaches based on the anonymized scenario: no actual data is involved.
Queries executed inside your VPC
NeuBird AI executes the actual telemetry queries locally within your VPC, against your real data, using the guidance from step 2.
Problem resolved, data stays internal
The ITOps problem is resolved using AI intelligence combined with your real data, none of which ever left your secure environment.