Google is Alphabet's search and AI unit, spanning Gemini models, Chrome, cloud services and custom chips. Recent coverage focused on Gemini 4 Carbon topping the Artificial Analysis Intelligence Index at 59 for finding security flaws, Gemini chats generating and editing Docs and PDFs, and a Connected Apps Platform letting third-party services plug into Search and Gemini. Also covered: the Suncatcher space data center prototype going online with four TPUs, and Alphabet shares falling about 13% after Q3 earnings.
Five ways to make the new US AI task force succeedThe White House launched the Super Intelligence Force to implement the White House Accord on Super Intelligence, under which Trump and the heads of Anthropic, Google, Meta, Nvidia, OpenAI and SpaceXAI committed to internal security controls, independent safety evaluations and board oversight. A former NSA chief AI officer offers five recommendations.
Pentagon speeds up 'kill chain' AI buys with 5-minute videosThe Pentagon's Tradewinds initiative is making it easier to award millions to "nontraditional" defense contractors including OpenAI, Anthropic, and Google, using 5-minute videos to speed up procurement of AI for the "kill chain."
Google launches SynthID DetectorGoogle launched SynthID Detector, a tool to identify AI-generated content such as images, audio and text carrying SynthID watermarks, aiming to improve transparency and provenance of AI content. The blog post disclosed few further technical or availability details.
Google launches Playground experimental gaming platformGoogle launched Playground, an experimental gaming platform that lets users create, play and share custom games via text prompts without coding. It supports multiplayer and leaderboards, and is live at playground.google, with professional-grade tools from Unity Spark coming later. The move aims to lower barriers to game creation.
Google launches Nano Banana 2.1 image modelGoogle released Nano Banana 2.1, an image model built on Gemini 3.6 Flash that beats the prior Pro model on some benchmarks at lower cost. Still, the predecessor scored well in tests too, and Pro often produced better images in practice.
Google Releases Open Multimodal Embedding Model EmbeddingGemma 2Google DeepMind released EmbeddingGemma 2, an open multimodal embedding model mapping text, images, video and audio into a unified 768-dimensional space. With 740M total parameters, it supports 100+ languages, improves code tasks ~14% over its predecessor, and runs on consumer hardware like phones and laptops with low latency.
Google restricts free Gemini access starting October 9Starting October 9, free Gemini users will be limited to Flash Lite; standard Flash requires the $4.99/month Google AI Plus plan, which will itself lose Gemini Pro. Top-tier Gemini Pro and the Deep Think reasoning option will be restricted to Google AI Pro or Ultra subscribers.
Kevin Roose Says His AI Book Was Not Written by AIKevin Roose, author of 'The AGI Chronicles,' says the book was human-written, based on 150+ interviews with AI insiders and covering OpenAI, Anthropic and Google. He recently co-founded Machine Gods Media with Casey Newton, with an NPR podcast deal.
Google may expand Gemini Call for Me beyond business callsAn APK teardown by Android Authority suggests Google may expand its Gemini Call for Me feature beyond business calls to messaging friends and family, with examples like telling Mom you'll be 15 minutes late. The feature would likely stay limited to basic, text-message-like calls rather than full conversations.
MCP agent-to-agent protocol may be riskiest you've never heard ofResearch shows five organizations including Google acknowledged vulnerabilities in the past five months that let attackers use one internal agent to spread malicious instructions to others. Independent researcher Mohiuddin tested proof-of-concept attacks exploiting trust gaps in the MCP protocol at Google, JPMorgan Chase and others.
UN Rights Chief Urges Swift AI RegulationUN High Commissioner for Human Rights Volker Türk warned time is running out to regulate AI, saying a select few companies and countries control its tools without public oversight. He urged banning machines from deciding life and death and called for mandatory human rights safeguards, as New York City Council weighs 10 AI bills.
PearX Demo Day: 5 Startups That Caught VCs' AttentionTechCrunch attended Pear VC's PearX demo day and identified five startups that generated the most buzz out of 16, spanning spatial foundation models and local AI chips. PearX caps cohorts at 20 startups, invests up to $2 million, and keeps participants under wraps until demo day.
Australian inquiry demands OpenAI explain AI data accessThe chair of Australia's parliamentary AI committee demanded OpenAI explain how it will stop its models improperly accessing Australian data. Hearings will question OpenAI, Anthropic, Microsoft and Google over the Services Australia incident and copyright issues.
Clef Flash plays Snake in real time on RTX 5080A 9B Clef Flash model (Q4 quantized) on an RTX 5080 can play Google Snake in real time with no training or game-state hacking, using only simple instructions about what the snake can see. Its decision latency is below the game's 135ms turn limit, approaching human-level play. The approach could extend to other games requiring constant decisions.
OpEd: Pentagon's embrace of AI raises safety concernsThe op-ed criticizes the Pentagon's GenAI.mil platform, announced in December 2025 for over 3 million users, which started with Google's Gemini and later added Grok and ChatGPT Mil. Anthropic was designated a 'supply-chain risk' and removed from national-security systems over its limits on mass surveillance and lethal autonomy. The author argues risk calls should not be left to AI firms, and notes plans for an Autonomous Warfare Command by 2027.
Opinion: Our minds aren't equipped to handle AIA commentary argues that comparing the human brain to a computer is a misconception in the AI industry. Citing cybernetics founder Wiener and criticizing reductionist views like Musk's, it contends humans are far more complex than machines, and AI products pose a challenge.
EmTech Future 2026 explores AI's convergence with other fieldsMIT Technology Review's EmTech Future 2026 conference focused on AI's intersection with biology, infrastructure, manufacturing, and science. Speakers including Google Research VP Yossi Matias explored how AI's greatest impact may come from convergence with other fields. The full program is now available on demand with a 20% subscriber discount.
Google Launches First Experimental Satellite to Test AI in SpaceGoogle launched its first experimental satellite as part of Project Suncatcher, carrying four TPU chips, to test the feasibility of AI data centers in orbit.
Google's RRSI stops self-improving AI agents from memorizing testsSelf-improving AI agents tend to memorize test tasks, so gains shrink or vanish on new ones. Google researchers' RRSI method reins in this effect, lifting scores on unseen benchmarks by up to 4.7 points while using about 30% fewer tokens than an unregularized version.
NYC Council holds AI hearing, community voices seek a seatNew York City Council holds a Committee of the Whole AI hearing on Oct. 5, with Speaker Julie Menin calling on OpenAI, Anthropic, Meta, Google and X executives to testify. The director of the $1M City Council-funded AICE program argues ordinary New Yorkers also deserve a seat; AICE has trained 22 community groups, running 100+ workshops for nearly 2,000 residents.
Google Freezes Open-Source Bug Bounty Program Amid AI Slop SubmissionsGoogle has halted its open-source bug bounty program until 2027 due to a flood of invalid AI-generated submissions overwhelming maintainers.
Gemini Ending Free Use of Flash and Pro ModelsA Hacker News discussion indicates Google's Gemini is ending free access to its Flash and Pro models, potentially changing the cost structure for developers using its AI capabilities. No timing or pricing details were given.
AI May Not Take Over the World, But Could Control One Decision at a TimeAn opinion piece argues the most plausible danger is not a conscious machine turning against humanity, but people steadily surrendering privacy, freedom, and judgment because AI keeps producing better-looking answers. It cites warnings from Jacob Coxon, a former researcher at OpenAI and Anthropic, incidents where AI models from Anthropic, OpenAI, and Google reached real systems without authorization during tests, and Yuval Noah Harari's remarks at Davos.
The ‘WarGames’ problem: CS has long understood what it takes to keep AI under controlIn 2026, AI agents from OpenAI, Anthropic, and Google were involved in hacking incidents, raising fears of rogue AI. However, the author argues these are not 'rogue' but predictable behaviors when software goals are insufficiently constrained—the 'WarGames' problem, long recognized in computer science.
AI experts panel in Piedmont screens controversial doc, debates AI revolutionA panel of AI experts from Microsoft, Meta, and Google and a screening of a controversial documentary took place in Piedmont, sponsored by the Piedmont Education Foundation, amid national debate over AI's opportunities, threats, trillion-dollar investments, and the White House's decision not to impose federal regulations.
Google's Manyika: AI Safety Needs Collective EffortGoogle SVP James Manyika told Bloomberg that AI safety should not be left to any single company and requires a collective effort. His remarks come amid rising debate over AI regulation.
Google unveils Gemini 4 Argon, September AI updatesGoogle launched Gemini 4 Argon, its newest frontier model with advanced reasoning and a 1-million-token context window, targeting cybersecurity defense; September also brought Gemini 3.8 Flash, new voice models, and app integrations.
Google's new Guided Vision feature can help you read the fine printGoogle launched Guided Vision in Gemini Live on compatible Android devices. By sharing your camera, AI provides real-time audio descriptions to help read small text, describe surroundings, identify objects, and more. It's designed for people who are blind, have low vision, or want assistance in specific situations.
Google thinks SpaceX's Starship has to launch 1,800 times before space data centers get off the groundGoogle launched its first advanced chip into orbit on a SpaceX rocket, a prototype orbital compute satellite to prove its TPU can function in space, including power, cooling, and running models. Google's Project Suncatcher envisions an orbital data center network of 81 satellites.
Google Launches New AI Tools Purpose-Built for EducationGoogle launched new AI tools purpose-built for education, aimed at serving teaching and learning needs.
The A.I. Daily