Google DeepMind Releases 10-Meter Global Forest Map ForTy 2020 as 1Password Debuts AI Agent Credential Broker

Google DeepMind and Google Research, working in partnership with the World Resources Institute and the International Institute for Applied Systems Analysis, released ForTy 2020 on Thursday, October 22, 2026, delivering a global geospatial dataset and open research program that maps every 10-by-10-meter patch of land on Earth across six forest categories. Published on the Google DeepMind blog, the multimodal satellite model tracks annual land-use transitions from 2017 through 2025.

Unlike conventional tree-canopy maps that treat commercial monoculture timber lots the same as ancient biodiversity reserves, ForTy 2020 distinguishes between primary forest, naturally regenerating forest, planted forest, plantation forest, tree crops and agroforestry, and non-forest land. Independent validation committees confirmed a 95 percent accuracy rate separating forest from non-forest terrain and a 90 percent accuracy rate across the six specific forest classifications, giving climate researchers and public environmental agencies in Silicon Valley, San Francisco, and San Jose high-precision compliance data for international deforestation regulations.

Across the enterprise software stack, password manager 1Password debuted a security architecture called 1Password for Agents that allows autonomous AI assistants—including Claude Duet, ChatGPT Work, Dots, Meta Muse, Grok Bot, and Town—to authenticate into web services without exposing plaintext credentials to the underlying language model. Chief Executive Officer Aashish Tripathi told The Verge that human operators must explicitly approve each credential injection request, preventing prompt-injection leaks across corporate workflows.

Meanwhile, a sobering computational biology benchmark highlighted the gap between in-silico model generation and wet-lab therapeutic efficacy. As detailed by The Verge, Citrini Research and San Francisco-based Anthropic spent approximately $50,000 in compute and grant funding using an unreleased Claude model alongside Twist Biosciences and Adaptyv Bio to screen nearly 320,000 candidate proteins for scorpion antivenom, generating promising initial binders in laboratory assays but failing to produce a clinically viable antivenom.