Agent Constitution: Policy Enforcement and PII Protection for AI Agents

AI agents are getting more capable. They can browse the web, call APIs, read and write files, and execute code. That capability is exactly what makes them useful and exactly what makes them dangerous