Google AI Updates: October 23, 2026

1. Google DeepMind Releases ForTy, a Global 10-Meter Map of Forest Types

Google. Google DeepMind, Google Research, the World Resources Institute, and IIASA published the Forest Typology (ForTy) 2020 map, which classifies every 10 by 10 meter patch of land into six classes (primary, naturally regenerating, planted, and plantation forest, tree crops and agroforestry, and other land), plus an early-release annual time series for 2017 to 2025. The model uses AlphaEarth Foundations satellite embeddings and about 1.7 million labeled locations; validation at more than 9,000 random sites found it identified the correct forest type 90% of the time and separated forest from non-forest 95% of the time, and it outputs per-class probabilities. The data is freely available on Google Earth Engine under an open license, with the 2020 map also on Figshare and a paper in Scientific Data. Source

2. Google Publishes Project Chrysalis, a Reference Architecture for Self-Evolving SRE Agents

Google. A Google for Developers post describes Project Chrysalis, a research reference architecture in which an ADK-based agent running Gemini 4 Argon diagnoses a production outage, writes and mounts a mitigation tool, scores its actions, stores lessons in memory, and proposes changes to its own structure that require signed human approval. It layers four security gates: Model Armor and an AST sandbox on inputs and generated code, Cloud KMS signing with Binary Authorization for patches, OPA policy-as-code with dual-custody approval, and tamper-evident Cloud Logging. In a simulated Black Friday cache stampede, Google reports time to first mitigation dropping to 1.8 seconds and P99 latency falling from 519.8 ms to 18.2 ms. A reference implementation is in the GoogleCloudPlatform/generative-ai repository under agents/self-evolving-ai. Source