Google: latest AI news
All Oossa news about Google, newest first.
Known for: Gemini, Gemma, DeepMind
Follow via RSSToday
Tools & agents
Google API Gateway adds streaming support in public preview
Developers can now run SSE, WebSocket, and gRPC streams behind API Gateway without changing their OpenAPI specs.
Models
Google DeepMind releases detailed global forest type map
The new Forest Typology 2020 map shows every 10‑meter patch of land as natural, planted or farmed forest, with yearly changes from 2017‑2025.
Business
Google TPUs enter Meta’s production AI systems
Meta is running AI workloads on Google’s custom chips through PyTorch-native versions of two inference engines. The announcement gives few details on scale or availability.
Models
Gemma 4 speech encoder evaluated on Saudi and Gulf Arabic without weight changes
Researchers tested a frozen Gemma 4 audio encoder and a lightweight adapter version on Arabic call‑center audio, finding comparable error rates to larger commercial models.
Yesterday
Safety
OpenAI and Google AI models spot critical PHP bug affecting many sites
GPT‑5.6 Sol and Gemini 4 Argon identified a severe PHP flaw; a GitHub tool now lets operators test their servers.
Models
Unstructured launches Transform models that out‑perform GPT‑6, Claude and Gemini
In tests on 224 business pages, Unstructured's new Transform models made far fewer table errors and hallucinations than leading AI models.
Models
Google adds Gemini voice navigation to Apple CarPlay
Google's Gemini AI can now answer spoken navigation questions in Google Maps on CarPlay, starting in the United States.
Policy & law
US bans Deepseek AI products; China says it’s discrimination
The US Department of Homeland Security ordered federal agencies to drop Deepseek tools, while China blocks ChatGPT, Claude and Gemini at home.
Business
OpenAI adds Google Cloud to its AI model lineup
OpenAI will let developers run its GPT models on Google Cloud, joining Microsoft and Amazon as hosts.
Business
Google Cloud adds NVIDIA Vera Rubin A5X instances for AI training and inference
Google Cloud will make NVIDIA Vera Rubin‑based A5X instances generally available next month, with plans to expand to Europe and scale up to 960,000 Rubin GPUs.
Tools & agents
Google launches SynthID Detector for watermark detection across multiple AI providers
The tool lets anyone upload media to check if it carries a watermark from Google, OpenAI, NVIDIA, Kakao and soon Apple.
Models
llama.cpp v b11772 adds Gemma‑4 fix and new binaries
The open‑source LLM runtime released binaries for macOS, iOS and several Ubuntu builds, plus a bug fix for Gemma‑4 models.
Tuesday, October 20
Models
Google adds Gemini Personal Intelligence to Android Auto
Google starts rolling out Gemini’s Personal Intelligence feature to Android Auto on Oct. 20, 2026, with broader car integration planned.
Business
Hugging Face releases TorchTPU for native PyTorch on Google Cloud TPUs
The new TorchTPU package lets PyTorch code run directly on Google Cloud TPUs, removing the need for torch_xla or Optimum TPU.
Models
Google adds native PyTorch support for vLLM and SGLang on TPU
Google launches PyTorch backends for the vLLM and SGLang inference engines, letting developers run these models on TPU hardware with familiar PyTorch tools.
Tools & agents
DeepMind agents finish Minecraft in 27 hours using self‑written code
Four Gemini 4 Argon agents wrote 2,783 programs, died 605 times and spent $520 in compute to kill the Ender Dragon.
Monday, October 19
Business
Amazon SageMaker Studio now reachable via simple AWS sign‑up
New developers can open a ready‑to‑use SageMaker Studio notebook in about 20 seconds using Google, GitHub, Apple or Amazon accounts, and get $100 free credits.
Models
Google releases LiteRT Tensor API for native edge model authoring
The new API lets developers hand‑craft C++ model graphs that run on CPU, GPU or NPU without rewriting code for each hardware.
Business
Google to finance extra wind, solar and batteries for Minnesota data center
Google will pay for 1,400 MW wind, 200 MW solar and 300 MW storage to support a 780 MW AI data center in Pine Island, Minnesota.
Tools & agents
Accenture Federal Services and Google launch federal Gemini program
The companies announced Gemini launchpads to help federal agencies move from agent experiments toward mission use at scale. Details are limited in the announcement excerpt.
Models
Google’s EmbeddingGemma 2 Adds Multimodal Search
Google’s open-weight model maps text, code, images, video and audio into one embedding space, with smaller versions for on-device use.
Models
Mistral AI launches Timestral, a 209‑million‑parameter forecasting model
Mistral AI unveiled Timestral, a family of time‑series models from 5 M to 209 M parameters that beats Google’s TimesFM‑3 on several benchmarks and runs on a CPU.
Sunday, October 18
Models
Google's Gemini 4 agents defeat Minecraft's Ender Dragon autonomously
Four Gemini 4 Argon AI agents explored a random Minecraft world, coordinated themselves, and killed the Ender Dragon without any human direction.
Models
Google Research Releases Harness2 Under Apache-2.0
Harness2 can rewrite an AI agent’s harness at test time. MarkTechPost reports an 11.5-point WorkBuddy-Bench gain for GLM-5 after one refinement step, without changing its weights.
Friday, October 16
Business
Replit adds Bring Your Own Cloud options for enterprises
Replit now lets companies run the full platform or just published apps in their own Azure, Google Cloud, or AWS tenant.
Models
Google releases EnvHarness for self‑evolving AI agents
The open‑source framework lets developers tweak training environments on the fly, boosting agent performance by up to 3.7 % on several benchmarks.
Models
Google's Gemini 4 Argon hits 97.5% on ARC‑AGI‑3 benchmark
Gemini 4 Argon scored 97.5% on the toughest general‑problem‑solving test, second only to OpenAI's GPT‑6 Astra.
Tools & agents
NousResearch releases standalone Solstice plugin for Hermes Agent
Version 0.2.0 adds Gemini 4 Argon and lets users run Hermes with their own per‑user Gemini quota via OAuth.
Thursday, October 15
Policy & law
Google detector flags Trump’s White House column photo as AI-edited
A Truth Social photo comparing a White House column before and after restoration was flagged by Google’s SynthID Detector. The White House had not commented.
Business
Amazon SageMaker Studio adds Google, GitHub, Apple sign‑in
New sign‑in options let users start with a ready‑to‑use Studio environment and free AWS credits, no traditional AWS account needed.