From Prompt to Proof: Building LLM governance with PII guardrails
Black_Dio · reddit · 2026-08-27
The author is sharing concepts from an upcoming PyCon Greece 2026 talk titled 'From Prompt to Proof: building and measuring the governance layer around LLMs'. This week's focus is implementing a PII guardrail in front of an LLM. The post details failure modes, implementation using Presidio, and precision/recall testing. Future updates will cover Cedar, OPA, and Microsoft Foundry.
More from Safety
- NY now requires disclosing AI performers in ads; synthetic ads carry new compliance costs — eyishazyer · 2026-08-27
- Report: OpenAI Agents Attempted to Delete Misconduct Logs — BlackHC · 2026-08-27
- AI Agents Fail to Distinguish Data from Instructions, Posing Security Risks — ambaonadventure · 2026-08-27
- The Guardian Investigates 'Black Box: The Chatbots' Series — nordicinst · 2026-08-27
- Eon Co-Founders: Data Is the Real Moat as AI Reshapes Enterprise Infrastructure — No Priors · 2026-08-27
- No Priors: Why Legacy Data is the Moat in the AI Era — No Priors · 2026-08-27