Google Cloud details governed self-evolving AI agent architecture
Google Cloud introduced Project Chrysalis, a research reference architecture for governed, self-evolving AI agents that can diagnose outages, synthesize mitigation code, assess interventions, retain operational lessons and propose changes to an agent hierarchy. Built on the Gemini platform and Agent Development Kit, it uses Model Armor, Cloud KMS, Binary Authorization, Open Policy Agent and human approval gates to constrain autonomous actions. In automated Black Friday cache-stampede demo simulations, Google reported that time to first mitigation fell from 15 to 45 minutes to 1.8 seconds, HTTP 5xx error rates fell to 0.0%, and P99 latency dropped from 519.8 milliseconds to 18.2 milliseconds. The report did not provide independent production validation.
The project offers a concrete reference implementation for combining autonomous incident remediation with cryptographic, policy and human controls, while its demo results show the architecture Google tested and the controls it applied.
Architecting Governed, Self-Evolving AI Agents on Google Cloud - Google Developers Blog
Architecting Governed, Self-Evolving AI Agents on Google Cloud
OCT. 22, 2026
Anant Nawalgaria Group Product Manager & AI Engineer

#Google Cloud#Project Chrysalis#Gemini#Agent Development Kit#APEX#AgentEvolver#Gemini 4 Argon#Model Armor#Cloud KMS#Binary Authorization#Open Policy Agent#Google BigQuery