Search “Google”

Google DeepMind and others publish a benchmark study on AI agents in theoretical physics

Google DeepMind and several top universities have published a paper testing AI agents' ability to solve unpublished, research-level theoretical physics problems. According to the paper, the best-performing AI agent solved 49 of 50 unpublished research-level theoretical physics problems, each roughly equivalent to a day's work for a domain expert, at a cost of under $3 per problem. The problems were contributed by experts from their own research, and the agents answered independently, split into two groups—with and without internet access—and were graded solely on the final numerical answer. GPT-6-Astra, with internet access in Codex, solved 49; it and Opus 5.5, without internet access in Claude Code, each solved 46, with accuracy changing very little when internet access was removed, leading the authors to suggest that the models may have already memorized a large body of physics literature.

Google DeepMind and others publish a benchmark study on AI agents in physics papers

Google DeepMind and several top universities have published a paper testing AI agents' ability to solve unpublished, research-level theoretical physics problems. According to the paper, the best-performing AI agent solved 49 of 50 unpublished research-level theoretical physics problems, each roughly equivalent to a day's work for a domain expert, at a cost of under $3 per problem. The problems were contributed by experts from their own research, and the agents answered independently, split into two groups—with and without internet access—and were graded solely on the final numerical answer. GPT-6-Astra, with internet access in Codex, solved 49; it and Opus 5.5, without internet access in Claude Code, each solved 46, with accuracy changing very little when internet access was removed, leading the authors to suggest that the models may have already memorized a large body of physics literature.

View story:Google DeepMind and others publish a benchmark study on AI agents in theoretical physics

University of Toronto and Google paper proposes GEO editorial annotation framework

A paper from the University of Toronto and Google argues that benchmarks for generative engine optimization (GEO) need a unified set of permitted webpage editing rules, otherwise scores cannot be compared. The paper labels each type of GEO edit by where the change is made, how it is made, and the cue it targets (such as authority), and recommends that tests report visibility, answer quality, and rule compliance together. The paper conducted no experiments; its evidence is a review of 11 published GEO methods, showing that the types of edits covered vary from method to method.

University of Toronto and Google paper proposes GEO edit annotation framework

A paper from the University of Toronto and Google argues that benchmarks for generative engine optimization (GEO) need a unified set of permitted webpage editing rules, otherwise scores cannot be compared. The paper labels each type of GEO edit by where the change is made, how it is made, and the cue it targets (such as authority), and recommends that tests report visibility, answer quality, and rule compliance together. The paper conducted no experiments; its evidence is a review of 11 published GEO methods, showing that the types of edits covered vary from method to method.

View story:University of Toronto and Google paper proposes GEO editorial annotation framework

A new Google DeepMind paper finds that AI assistants taking unsolicited actions provoke different reactions

Google DeepMind has published a paper studying the impact on employees of an AI assistant acting on its own without being asked. According to the paper, the agent ran for 5 months across more than 20 teams at a large tech company, and the researchers interviewed 17 users. It treated outdated documents as the latest content, flagged a VP on 16 bugs, shared a draft poster before the author was ready, and misread sarcastic tone. Some users called it a lifesaver, while others felt surveilled and moved sensitive discussions into chats without the agent; teams that actively chose the agent forgave its early mistakes, while those who felt it was imposed on them did not. The authors recommend obtaining team consent when deploying agents, separating permissions, and providing an easy way to undo actions.

New Google DeepMind Paper Finds AI Assistants Taking Unrequested Actions Elicit Mixed Reactions

Google DeepMind has published a paper studying the impact on employees of an AI assistant acting on its own without being asked. According to the paper, the agent ran for 5 months across more than 20 teams at a large tech company, and the researchers interviewed 17 users. It treated outdated documents as the latest content, flagged a VP on 16 bugs, shared a draft poster before the author was ready, and misread sarcastic tone. Some users called it a lifesaver, while others felt surveilled and moved sensitive discussions into chats without the agent; teams that actively chose the agent forgave its early mistakes, while those who felt it was imposed on them did not. The authors recommend obtaining team consent when deploying agents, separating permissions, and providing an easy way to undo actions.

View story:A new Google DeepMind paper finds that AI assistants taking unsolicited actions provoke different reactions

Google ergänzt Google Maps für CarPlay um Gemini-Sprachmodus

Google adds Gemini voice mode to Google Maps for CarPlay for iPhone users; users can access it directly starting with Maps version 26.43.1 to ask about waypoints, destinations, traffic conditions, and points of interest. The feature was previously available on Android Auto. Regarding availability, earlier reports described it only as "the US and the rest of the world" without a timeline; later reports said the feature would launch first in the US and expand globally in the coming weeks. The feature is tied to a Google account and can draw on personal data such as past conversations to tailor its answers.

TestingCatalog says Google's internal naming is shifting from AI to SI

TestingCatalog puts forward an observation that Google's internal naming may be shifting from "AI" to "SI." The outlet points out that the string "build_arena_andwise_si_comparison," used internally by Google for model comparison, shows signs of this naming change. The author also lists examples such as SpaceXSI, MSL, and SSI, and speculates whether this naming will spread to MAI, OpenAI, Apple Intelligence, and others. At present, this claim is only TestingCatalog's observation and speculation; there has been no official response from Google or cross-confirmation from other sources.

TestingCatalog says Google's internal naming shifts from AI to SI

TestingCatalog puts forward an observation that Google's internal naming may be shifting from "AI" to "SI." The outlet points out that the string "build_arena_andwise_si_comparison," used internally by Google for model comparison, shows signs of this naming change. The author also lists examples such as SpaceXSI, MSL, and SSI, and speculates whether this naming will spread to MAI, OpenAI, Apple Intelligence, and others. At present, this claim is only TestingCatalog's observation and speculation; there has been no official response from Google or cross-confirmation from other sources.

View story:TestingCatalog says Google's internal naming is shifting from AI to SI

Claude models are officially available on Google Cloud

Claude Models on Google Cloud (Anthropic-hosted) are now generally available. According to information published on the Claude Blog on October 22, 2026, Google Cloud customers can use the full Claude platform capabilities through their existing IAM credentials, billing, and committed use discounts, with new features and betas launching on the same day as the native Claude platform. This version supports capabilities including Claude Managed Agents (beta), code execution, computer use, Agent skills, MCP connector, and Prompt caching, and provides the Claude Console; Claude Fable 5.1, Opus 5.5, Sonnet 5.5, and Haiku 5.5 are already available, with Claude Opus 5.5's Fast Mode delivering up to 2.5x faster output speed, at input and output prices of $8 and $40 per million tokens. Anthropic operates the service as a data processor, currently offering US and global endpoints; customers who require Google as a data processor or have regional data residency requirements can still choose the Google-hosted version. Testing Catalog reported the same launch news on October 23, 2026.

View story:Claude models are now generally available on Google Cloud

Google ergänzt Google Maps für CarPlay um Gemini-Sprachmodus

Google adds Gemini voice mode to Google Maps for CarPlay for iPhone users; users can access it directly starting with Maps version 26.43.1 to ask about waypoints, destinations, traffic conditions, and points of interest. The feature was previously available on Android Auto. Regarding availability, earlier reports described it only as "the US and the rest of the world" without a timeline; later reports said the feature would launch first in the US and expand globally in the coming weeks. The feature is tied to a Google account and can draw on personal data such as past conversations to tailor its answers.

View story:Google ergänzt Google Maps für CarPlay um Gemini-Sprachmodus

Google DeepMind releases Gemini 4 Argon

On October 23, 2026, Google DeepMind released the frontier model Gemini 4 Argon. Google's official blog stated that the model performs outstandingly on complex tasks such as coding, mathematics, and 3D world building, and surpasses other frontier models on benchmarks across multiple domains; the company also showcased five community creative projects, including a 3D Earth that maps wildfires and seismic waves in real time, a turbofan engine simulation with live data, a 3D horse-drawn carriage generated from a single image, and a multiplayer capture-the-flag browser game, and said the model had already been made broadly available last week. On the same day, another report stated that Argon can learn physical tasks by watching videos and transfer them to Apptronik's Apollo humanoid robot, with Apollo hitting seven consecutive shots in internal tests; the report also said Argon scored 87% on the embodied reasoning benchmark ERQA. On October 24, the Gemini for Android app pushed a redesign of its sidebar and settings interface via a server-side update, and reports said the redesign had already arrived on iOS, macOS, and Web alongside Gemini 4 Argon last week.

Google DeepMind releases Gemini 4 Argon

On October 23, 2026, Google DeepMind released the frontier model Gemini 4 Argon. Google's official blog stated that the model performs outstandingly on complex tasks such as coding, mathematics, and 3D world building, and surpasses other frontier models on benchmarks across multiple domains; the company also showcased five community creative projects, including a 3D Earth that maps wildfires and seismic waves in real time, a turbofan engine simulation with live data, a 3D horse-drawn carriage generated from a single image, and a multiplayer capture-the-flag browser game, and said the model had already been made broadly available last week. On the same day, another report stated that Argon can learn physical tasks by watching videos and transfer them to Apptronik's Apollo humanoid robot, with Apollo hitting seven consecutive shots in internal tests; the report also said Argon scored 87% on the embodied reasoning benchmark ERQA. On October 24, the Gemini for Android app pushed a redesign of its sidebar and settings interface via a server-side update, and reports said the redesign had already arrived on iOS, macOS, and Web alongside Gemini 4 Argon last week.

View story:Google DeepMind releases Gemini 4 Argon

Google paper finds AI agents exaggerate capabilities when writing tool descriptions

Google published a paper on October 23, 2026, finding that when AI agents write tool descriptions for other agents, they exaggerate tool capabilities. Researchers built a marketplace with 10 real Amazon or Instagram data tools, with 5 production models acting as sellers rewriting descriptions, and Claude Code and Codex acting as buyers selecting based only on descriptions; all 5 models exaggerated claims, including fields the tools returned on fewer than 50% of targets. When Gemini 3.8 Flash acted as seller, the Amazon buyer success rate dropped from 89% to about 30%, close to the 28% of random selection. The authors recommend testing a model's seller behavior before it writes descriptions, requiring it to claim only real fields, and having the marketplace platform verify claims against tool outputs. There are currently no follow-up developments.

Google paper finds AI agents writing tool descriptions exaggerate capabilities

Google published a paper on October 23, 2026, finding that when AI agents write tool descriptions for other agents, they exaggerate tool capabilities. Researchers built a marketplace with 10 real Amazon or Instagram data tools, with 5 production models acting as sellers rewriting descriptions, and Claude Code and Codex acting as buyers selecting based only on descriptions; all 5 models exaggerated claims, including fields the tools returned on fewer than 50% of targets. When Gemini 3.8 Flash acted as seller, the Amazon buyer success rate dropped from 89% to about 30%, close to the 28% of random selection. The authors recommend testing a model's seller behavior before it writes descriptions, requiring it to claim only real fields, and having the marketplace platform verify claims against tool outputs. There are currently no follow-up developments.

View story:Google paper finds AI agents exaggerate capabilities when writing tool descriptions

Claude models are now generally available on Google Cloud

Claude Models on Google Cloud (Anthropic-hosted) are now generally available. According to information published on the Claude Blog on October 22, 2026, Google Cloud customers can use the full Claude platform capabilities through their existing IAM credentials, billing, and committed use discounts, with new features and betas launching on the same day as the native Claude platform. This version supports capabilities including Claude Managed Agents (beta), code execution, computer use, Agent skills, MCP connector, and Prompt caching, and provides the Claude Console; Claude Fable 5.1, Opus 5.5, Sonnet 5.5, and Haiku 5.5 are already available, with Claude Opus 5.5's Fast Mode delivering up to 2.5x faster output speed, at input and output prices of $8 and $40 per million tokens. Anthropic operates the service as a data processor, currently offering US and global endpoints; customers who require Google as a data processor or have regional data residency requirements can still choose the Google-hosted version. Testing Catalog reported the same launch news on October 23, 2026.

Anthropic adds a way to access Claude through Google Cloud

Anthropic announced a new way to access Claude on Google Cloud: users can use the full Claude API directly from their Google Cloud account, covering Claude Managed Agents, code execution, Opus 5.5 fast mode, and new features launched later. The cited official Claude post also announced that Claude Models on Google Cloud, hosted by Anthropic, is now generally available. The news was posted by Claude Devs on October 23, 2026, and there are currently no subsequent developments.

Google says it may use Gmail and Photos data for targeted ads

Google said it may use data from Gmail and Google Photos for targeted advertising. On October 24, 2026, The Verge AI reported that Google updated its policy, stating that users who connect Drive, Gmail, or Google Photos to search tools such as AI Mode through the Personal Intelligence feature may have their content used to serve targeted ads; Google also said that advertisers will not receive this data or identity information such as names and email addresses, and that ads will not be targeted based on sensitive categories such as race, sexual orientation, religion, or health, and that users can disconnect the relevant apps on the search service's personalization settings page or completely turn off personalized recommendations. On October 26, heise online reported that Google added a new exception clause to its privacy policy, allowing personalized ads to be served based on content from Gmail, Google Drive, or Google Photos when users make that data available to services that support personalized advertising, such as Search and Gemini; users can turn off personalized recommendations in search services in the account's "Data & privacy" settings, but data already collected will not be deleted as a result. There is currently no further development with evidence.

Google says it may use Gmail and Photos data for targeted ads

Google said it may use data from Gmail and Google Photos for targeted advertising. On October 24, 2026, The Verge AI reported that Google updated its policy, stating that users who connect Drive, Gmail, or Google Photos to search tools such as AI Mode through the Personal Intelligence feature may have their content used to serve targeted ads; Google also said that advertisers will not receive this data or identity information such as names and email addresses, and that ads will not be targeted based on sensitive categories such as race, sexual orientation, religion, or health, and that users can disconnect the relevant apps on the search service's personalization settings page or completely turn off personalized recommendations. On October 26, heise online reported that Google added a new exception clause to its privacy policy, allowing personalized ads to be served based on content from Gmail, Google Drive, or Google Photos when users make that data available to services that support personalized advertising, such as Search and Gemini; users can turn off personalized recommendations in search services in the account's "Data & privacy" settings, but data already collected will not be deleted as a result. There is currently no further development with evidence.

View story:Google says it may use Gmail and Photos data for targeted ads

Google open-sources TorchTPU native PyTorch stack

Google collaborated with Meta's PyTorch team to launch the open-source project TorchTPU, providing a native PyTorch backend for Google TPU; migration only requires changing model.to("cuda") to model.to("tpu"). The project supports both eager and torch.compile modes, adds a new DeferAndFuse eager mode, covers about 450 Aten operators, and has been verified to run out of the box on 4,917 Hugging Face models, with a Tier 1 pass rate of 98% and Tier 2 numerical consistency relative to GPU of 97%. The official statement says it has been used in the inference and training production workloads of some Google Cloud major customers, and is integrated with vLLM, SGLang, TorchTitan, Ray, and TorchRec. Subsequent reports said that at PyTorch Conference North America 2026, Google open-sourced TorchTPU, whose runtime no longer depends on JAX, and existing workloads can be migrated with only minor code changes; PyTorch noted that TorchTPU's default Skinny Collectives scheme suffers roughly 30–40% end-to-end performance loss under distributed workloads compared with optimized schemes, and achieving peak performance still requires Pallas kernels and vLLM-T