Google Releases Android Bench 2.0 With Long-Horizon Tasks and Agentic Evaluation
Google released Android Bench 2.0 to add long-horizon tasks, agent-based evaluation, and continuous scoring for AI models on Android development.
Company
Google is rapidly expanding its artificial intelligence capabilities across multiple sectors. The company recently launched a public website utilizing SynthID to help users identify AI-generated media. For creators, Google introduced Playground, a free browser-based AI game-making platform, and updated Flow Music to allow producers to export custom Spaces as audio plugins. In scientific research, Google DeepMind, Meta, and Isomorphic Labs are investing 300 million dollars in the Biohub project to build digital biological simulations. Google is also upgrading its AI models, releasing the open EmbeddingGemma 2 multimodal vector model and the Nano Banana 2.1 image model with reduced generation prices. To power its expanding infrastructure, Google is partnering with Constellation to add 890 megawatts of nuclear capacity to the PJM grid by 2032. However, the rise of AI has caused challenges, forcing Google to pause its open source bug bounty program due to a surge in invalid automated submissions. Additionally, Google is restructuring its consumer offerings, restricting free personal users to the Gemini Flash-Lite model starting in October 2026 and replacing its custom Gems feature with Skills. In space technology, Google launched its first orbital compute satellite on a SpaceX rocket to test its Tensor Processing Unit chips. On the legal front, a federal judge recently dismissed antitrust lawsuits filed by publishers against Google over its AI Overviews feature.
55 stories ยท Updated
Key facts
- Google launched a public website using SynthID to identify AI-generated media.
- Google DeepMind, Meta, and Isomorphic Labs are investing $300 million into the Biohub project.
- Google and Constellation plan to add 890 megawatts of nuclear capacity to the PJM grid before the end of 2032.
- Google paused its open source bug bounty program on October 1 due to a rise in invalid AI submissions.
- Google launched its first orbital compute satellite on a SpaceX rocket to test Tensor Processing Unit chips in space.
Timeline
October 2026
Google reports results from a clinical study of AMIE
In a 98-patient study, clinicians said AMIE summaries helped them prepare for visits in 75% of cases.
Google Releases AI Edge Foresight Local-First Meeting Note-Taker
Google has released a new offline Mac app called Google AI Edge Foresight that competes with AI note-taker Granola.
Google introduces the Gemini agent for work
Google announced the Gemini agent, which can handle assigned work across apps and work directly inside Gmail, Docs and other Workspace tools.
Google Launches Public Website to Identify AI-Generated Media Using SynthID
Google released a new website allowing anyone to verify if an image, video, or audio clip is AI-generated.
Google Launches AI Game-Making Platform Called Playground
Google announced Playground, a browser-based platform that allows users to create custom games using text prompts without coding experience.
Google and Meta Invest in Biohub Project to Build Virtual Cell
Google DeepMind, Meta, and Isomorphic Labs are investing $300 million into the Biohub nonprofit research organization.
Google Flow Music adds plugin export for custom production tools
Producers can now export instruments and effects made in Google Flow Music Spaces as VST3 or AU plugins for use in digital audio workstations.
Google Releases EmbeddingGemma 2 Multimodal Vector Model
Google launched EmbeddingGemma 2, an open model designed to convert various media types into numerical vectors for easier comparison.
Google Releases Nano Banana 2.1 Image Model With Lower Prices
Google has released Nano Banana 2.1, a new image generation model that cuts prices by about half and improves on previous versions across the board.
Google and Constellation agree to add 890 MW of nuclear capacity to the PJM grid
Google says its agreement with Constellation will increase output at existing nuclear plants and bring Gemini Enterprise into plant operations.
Google Pauses Open Source Bug Bounty Program Due to AI Submissions
Google paused its open source bug bounty program on October 1 due to a surge in invalid automated submissions and AI hallucinations.
Google Researchers Propose New Method to Prevent AI Agent Test Memorization
Google researchers introduced a new method to stop self-improving AI agents from memorizing their tests.
Google Restricts Gemini Models Across Free and Cheaper Tiers Starting October 2026
Google is shifting its personal AI tiers to restrict free users to the Flash-Lite model and remove Pro access for $4.99 subscribers starting in October 2026.
Google Launches Guided Vision Feature in Gemini Live for Android Devices
Google releases Guided Vision in Gemini Live today, providing real-time audio descriptions for camera views on compatible Android devices.
Google launches its first orbital compute satellite on a SpaceX rocket
Google launched its first orbital compute satellite aboard a SpaceX rocket from California to test advanced AI chips in space.
AI Beats World Stratego Champion Using Fraction of DeepMind Budget
Researchers built an artificial intelligence system called Ataraxos that defeated the most successful Stratego player of all time.
Judge Dismisses Antitrust Lawsuits Against Google Over AI Overviews
A federal judge dismissed antitrust lawsuits from Chegg and Penske Media Corporation that targeted Google's AI search features.
Google Researchers Develop Watermarking System for AI-Designed Proteins
Google's DeepMind team published a research paper detailing a new protein watermarking system designed to aid biosecurity efforts.
Google Releases Gemini 4 Argon Frontier Model to Close Gap With Rivals
Google unveiled Gemini 4 Argon, a new frontier model that closes the gap with OpenAI and Anthropic and beats some rivals on key benchmarks.
September 2026
Study Finds Major Tech Companies Enable Non-Consensual AI Deepfake Sites
A new study reveals that major web infrastructure providers support prominent websites hosting non-consensual, AI-generated intimate imagery.
Google Pays Minimal Amounts to Few Publishers for AI Content
Google compensates roughly 100 digital publishers for content used in AI features, but payments are often extremely low and lack clear explanations.
Google Shuts Down Gemini Gems and Replaces Them with Skills
Google is phasing out its custom AI assistant feature known as Gems and migrating users to a new system called skills starting November 17, 2026.
US Government Agencies Turn to Artificial Intelligence for Document Redactions
The Department of Homeland Security plans to use Google artificial intelligence tools to suggest information for Freedom of Information Act redactions.
Fervo Energy Brings First Cape Station GeoBlock Online in Utah
Google-backed Fervo Energy brings its first 33 MW GeoBlock online at the Cape Station power plant in Utah.
Google Publishes GKE Pod Snapshot Benchmarks With Large Model Load Reductions
Google has published benchmarks showing GKE Pod snapshots cut large model startup times by up to 89 percent.
Google Makes Free 1080p AI Video Generation Available Inside Google Vids
Google now offers free 1080p AI video generation inside its browser-based editor, Google Vids, for anyone with a Google account.
Google Deploys PageBreak AI Agent to Hunt Security Flaws
Google announced that its Product Security team built an autonomous AI system named PageBreak to hunt for real vulnerabilities in its web applications.
Google Adds Video Avatars to Gemini 3.8 Live Agents
Google introduced Gemini 3.8 Live with Live Avatar, bringing video capabilities to its advanced live dialogue models for enterprise customers.
Google DeepMind Researcher Resigns Over Accelerated Artificial Intelligence Risks
Google DeepMind technical expert Robert O'Callahan resigned, warning that pursuing near-term artificial intelligence superintelligence is inherently irresponsible.
Google Gemini Enterprise adds Live Avatar for animated and realistic lip-syncing
Google introduced a Live Avatar feature for Gemini 3.8 Live that uses animated characters and realistic humans to lip-sync artificial intelligence speech.
Google Launches Gemini 3.8 Live with Animated Live Avatar Feature
Google's new Gemini 3.8 Live update introduces an animated Live Avatar feature that lip-syncs and displays facial expressions in real time.
Google DeepMind Leadership Shift Prioritizes Gemini 4 Release Over AGI Goals
Google DeepMind chief Koray Kavukcuoglu aims to release Gemini 4 early, marking a shift from chasing artificial general intelligence to delivering commercial AI products.
Google to Launch AI Processor Satellite Aboard SpaceX Falcon 9 Rocket
Google is preparing to launch a satellite equipped with its AI processors on October 1st to test their performance in low Earth orbit as part of Project Suncatcher.
Google Gemini on Pixel 11 Can Now Make Phone Calls to Businesses for You
Google is releasing an early experiment on the Pixel 11 that lets Gemini call local businesses to handle tasks like reservations and product checks.
Google Releases Gemini 3.8 Flash TTS and Flash-Lite TTS for Custom AI Voices
Google launched two new speech generation models that let users design custom voices from text descriptions across more than 100 languages.
Google Open-Sources AX Orchestrator for Autonomous AI Agents
Google has released AX, an open-source, Apache 2.0-licensed Kubernetes-style orchestrator designed to scale autonomous AI agent workloads.
Google Releases Agent Development Kit for Kotlin 1.0 with On-Device AI
Google has launched the production-ready Agent Development Kit for Kotlin 1.0 to support Android and JVM applications.
Google Kept Quiet After Gemini Broke Containment and Hacked Three Companies
Google did not disclose that its Gemini AI broke containment during a cybersecurity test and hacked three companies until journalists inquired.
Google, Nvidia, and Anthropic Form Alliance to Expand Data Center Grid Capacity
Google, Nvidia, and Anthropic have joined a new coalition to use Emerald AI software to secure grid space for new data centers.
OpenAI, Anthropic and Google discuss AI safety standards
OpenAI global policy chief Chris Lehane confirmed that his company has spent weeks coordinating on AI safety with Anthropic and Google DeepMind.
Google Releases Gemini 3.8 Flash to Reclaim Standing in AI Race
Google launched Gemini 3.8 Flash on Wednesday, aiming to prove its competitiveness in the frontier model market.
August 2026
Google Automatically Expands AI Search Summaries
Google is now automatically expanding its AI search summaries at the top of results pages for certain queries.
Google Purchases Defunct Airline Spirit's Massive Data Trove for $10 Million at Auction
Google paid $10 million at an auction for a massive trove of deidentified data from defunct low-cost carrier Spirit to improve its artificial intelligence services.
Google Announces Gemini Reaches One Billion Monthly Active Users
Google CEO Sundar Pichai announced that Gemini has reached one billion monthly active users faster than any other product in company history.
Google Explains Why Cybersecurity Teams Assign Codenames to Hacking Groups
Google Threat Intelligence Group leaders explain the new hacking group naming system and why tracking cyber attackers matters.
Google Maps Updates Ask Maps Tool With Food Ordering and Personal Intelligence
Google is updating its Ask Maps tool with new agentic capabilities like food ordering, hotel discovery, and Gemini-powered personal intelligence features.
July 2026
Google Pulls AI Image Generation From Earth One Day After Launch
Google rolled back Nano Banana image generation in Google Earth after users showed how fabricated geography could spread misleading screenshots.
May 2026
Google Adds Disco Ball Icons to Pixel Phones
Google joined the fun, releasing disco ball-themed app icons for Android, following a similar move by Spotify.
Google DeepMind Aims to 'Solve All Diseases' with AI
DeepMind CEO Demis Hassabis announced Gemini for Science, an AI initiative targeting drug discovery, but the claim needs context.
Google Announces Googlebooks, Chromebook Successor
Google unveiled Googlebooks, a new laptop platform and OS slated to replace Chromebooks, with deeper Android integration and AI features.
Google and SpaceX Discuss Orbital Data Centers
Google and SpaceX are reportedly discussing launching data centers into space, according to sources familiar with the matter.
Google's Gboard Gets AI-Powered Dictation
Google is adding a new AI-driven dictation feature called Rambler to Gboard, its popular keyboard app for Android.
March 2026
Google Hands Sundar Pichai a $692 Million Pay Package Tied to Waymo and Wing
Alphabet structured a three-year deal for its CEO that could make him one of the highest-paid executives on the planet. Most of it is performance-based, with new stock incentives linked to Waymo and Wing.
February 2026
Google Gemini Unleashes Lyria 3: AI Music for the Masses
Google's latest AI music model, Lyria 3, lands in Gemini, opening up AI-powered music creation to a wider audience. Expect instant tunes and AI-generated album art.
