Every published Top AI Stories item tagged with Google, newest first.
HEIR, short for homomorphic encryption intermediate representation, converts a pretrained model so that it computes on encrypted inputs, meaning the server holding the model never sees the underlying data. Google published the compiler on GitHub alongside four working demonstrations built with outside partners: a recommendation model, credit-card fraud detection, network intrusion detection and hotword detection, each timed on a single processor thread. Homomorphic encryption has been sound in theory and too slow in practice for years, and latency is the half this targets.
The energy research firm Noreva expects natural gas to pass $10 per million British thermal units at some delivery hubs in the coming years, against a Henry Hub price now just under three dollars. Meta has committed to a 7.5-gigawatt plant in Louisiana and Amazon to a 7.6-gigawatt plant in Texas, with Microsoft and Google building gigawatt-scale plants of their own. Chief executive Peter Gardett argues those buyers are carrying unusual price risk, which would surface either in token prices or on the wider grid.
A new Media Watermark toggle in settings lets users switch off the visible mark Google stamps on output from Gemini, the Nano Banana, Omni and Lyria models, and the Flow video editor, with Search support to follow. The invisible SynthID watermark and C2PA provenance metadata stay on either way, and Google open-sourced a library called Credentio so developers can validate C2PA credentials locally. The change arrives two weeks after the EU AI Act transparency duties and California's disclosure deadline became enforceable.
Google made Gemini 3.7 Flash generally available on August 13, calling it its most intelligent workhorse model yet for coding and agents. The jumps over 3.6 Flash are large for a three-week gap: long-horizon software engineering rises from 49.0 to 65.3 on DeepSWE version 1.1, and enterprise workflow automation nearly doubles from 17.0 to 30.4. Introductory pricing runs at 75 cents per million input tokens through December 31, then doubles on January 1.
A preprint posted August 10 finds that the encrypted reasoning blocks the major providers hand back to clients are interchangeable across sessions, users and models — so injecting a strong model's trace into a weaker, less-guarded one makes it print the hidden text in the clear. The authors decoded 315,320 publicly scraped blocks and recovered 367 pieces of personally identifiable information along with 182 credentials. They disclosed the flaw before publishing and propose cryptographic fixes for client-side reasoning.
Google DeepMind moved a sign-language-to-text model called SL2T out of research and into two shipping products, Gboard and Live Transcribe, on the Pixel 11 at no additional cost. It was trained on more than 100,000 hours across over 50 sign languages and reaches a zero-shot score of 70 on the BLEURT translation metric on the FLEURS-ASL benchmark, well above any previously reported result. Pose tracking runs on the device, so only geometric coordinates leave the phone. American Sign Language to English ships first, guided by an advisory committee of Deaf organizations.
Sundar Pichai said the Gemini app has passed one billion monthly users, making it the fastest product in Google's history to reach that mark and the fourteenth Google product to get there. The company had reported 950 million monthly users on its second-quarter earnings call weeks earlier. The figure counts the standalone app only, excluding people who meet Gemini inside Search, Workspace and Android — and it lands roughly two months after OpenAI's ChatGPT crossed the same threshold.
Google Research reported that AMIE (Video), a version of its medical dialogue system that handles live audio and video, was rated on par with board-certified primary care physicians on history-taking, diagnostic accuracy, management and communication. The randomized study covered 100 clinical scenarios and 300 consultations with trained patient actors, scored by 20 independent physicians. AMIE rated significantly higher than the doctors at eliciting physical signs and guiding examinations. It remains research-only — every consultation used actors, not real patients.
WeatherNext predicts a storm's track, intensity and wind structure together rather than trading one against the others, and its three-day forecasts now match what prior models managed at two days — a gain the team frames as roughly a decade of meteorological progress. It runs on a 28-kilometer grid and produces a forecast in under one minute on a single tensor processing unit. The National Hurricane Center, the UK Met Office and the Cooperative Institute for Research in the Atmosphere are already using it, and Google published the paper in Nature alongside open code and model weights.
In a message to staff, Sundar Pichai said Hassabis moves from chief executive of Google DeepMind to chair of the lab and chief scientist of Alphabet, handing day-to-day control to Koray Kavukcuoglu. Kavukcuoglu is promoted to senior vice president, reports directly to Pichai, and keeps his chief AI architect title while taking over Gemini model development, frontier research, and the Gemini app and developer teams. Hassabis continues to lead Isomorphic Labs and says he will concentrate on the path to artificial general intelligence.
Pichai's message also confirmed the exit of Jeff Dean, Google's 30th employee and the architect of much of its machine learning infrastructure. Dean is founding Discovery Loop with his longtime collaborator Sanjay Ghemawat plus Quoc Le and Oriol Vinyals, aiming to automate the experimental loop in science and engineering by running thousands of experiments in parallel, starting with machine learning research itself. Alphabet is a founding investor and cloud partner, and the first round is co-led by Radical Ventures and Khosla Ventures with Kleiner Perkins, Lightspeed and Doerr Capital joining.
High Bandwidth Flash sits between high-bandwidth memory and solid-state storage, using NAND flash to hold far more data than HBM does at speeds much closer to it. The first specification, published through the Open Compute Project six months after the consortium formed, defines stacks up to 512 gigabytes with three performance grades running from roughly 0.4 to 3 terabytes a second. The target is inference, where model weights and context windows have outgrown the memory attached to each accelerator. Google and Tenstorrent are also in the consortium.
The Financial Times detailed the structure behind Anthropic's five-year Google Cloud commitment, and it looks less like a cloud contract than like aircraft leasing. Morgan Stanley arranged a private-credit vehicle, Compute SPV, that buys Google's tensor processing units with outside money and leases them back to Anthropic — roughly $35 billion of hardware in the first tranche, about one million chips and one gigawatt of capacity, with Apollo and Blackstone anchoring the debt. Around $150 billion of the roughly $200 billion in linked contracts now rides on Anthropic making its lease payments.
The deadline set by Executive Order 14409 passed on August 1 with nothing published. The Office of the National Cyber Director has since invited OpenAI, Anthropic, Google and Meta to review a completed draft, which would let labs voluntarily submit frontier models for government testing of their hacking ability before release. Submission mechanics, testing standards and whether findings ever become public are all still unpublished — and because participation is voluntary, the labs reviewing the rules are also the ones who decide whether to follow them.
Executive Order 14409, signed June 2, gave federal agencies 60 days to stand up a classified benchmarking process, a voluntary framework for disclosing frontier models before release, and a plan to widen federal cybersecurity hiring. That clock ran out on August 1 with no Federal Register notice, nothing published by the National Institute of Standards and Technology or the Cybersecurity and Infrastructure Security Agency, and no statement from the Office of Science and Technology Policy. The framework would have given agencies a 30-day look at covered models before release — until it exists, labs are guessing which of their systems would even qualify.
Google said on July 31 that it will not ship the standalone AI Studio app for Android and iOS, despite roughly 800,000 preorders taken since the company teased it at its May developer conference. In its own words, rather than "asking you to download yet another app," Google is taking an approach "where apps emerge naturally, in the course of your everyday conversations with Gemini," and is working with the Gemini app team to bring prototyping to mobile and desktop. The AI Studio website continues for people going from an idea to a prompt to a business.
Google shipped a Google Earth feature that used its Nano Banana 2 image model to generate imagery and lay it over satellite maps, then withdrew it a day later. Critics noted that one of the most trusted sources of visual evidence for journalists and researchers had just been handed a fabrication tool. Google said it had seen people sharing generated imagery that appeared to violate its policies, and would roll the feature back while it builds stronger guardrails.
Across Chrome milestones 149 and 150, Google fixed 1,072 security bugs — more than it fixed across the previous twenty-three milestones combined. The work runs through a chain of Gemini-based agents: one hunts for vulnerabilities, one writes candidate fixes, a critic agent reviews them, and test-writing agents cover the result, alongside DeepMind's CodeMender and Project Zero's BigSleep. In May alone, continuous scanning blocked more than twenty vulnerabilities from reaching production rather than shipping and patching them later.
Reddit posted $805 million in second-quarter revenue, up 61 percent year over year, and net income of $253 million. The stock still fell more than 10 percent after hours, because US daily active uniques slipped to 53.2 million from 53.5 million and chief executive Steve Huffman said search referrals were choppy as Google's AI summaries answer questions that used to send readers to Reddit threads. The company also signaled that it may not renew its 2024 data-licensing deal with Google on the same terms.
Google put Lyria 3.5 into Flow Music, its consumer song-generation app, with richer melodic structure, closer adherence to lyric prompts, more emotionally nuanced vocals and better pronunciation, plus finer control over tempo and duration. It arrives as a free upgrade for existing Flow Music users rather than a new paid tier. Flow Music is Google's most direct answer to Suno and Udio, and Lyria is now the model underneath it.
New Similarweb data shows AI Overviews surfacing in 43 percent of United States searches, up from about 15 percent a year earlier, while monthly visits to Google's conversational AI Mode grew from 126 million to 279 million. Queries are getting longer and more conversational as users treat search as a destination rather than a doorway. For publishers that is the whole problem: answers that resolve in place generate no click, and only 6.8 percent of ChatGPT desktop replies carried a citation.
Nvidia chief executive Jensen Huang published a letter — his first-ever post on the social platform X — urging Washington not to restrict downloadable AI models, and within a day the signatory list doubled from 25 companies to 50, including OpenAI, Google, Microsoft, AMD, Cisco, and GitHub. The conspicuous absences were Anthropic, the most vocal proponent of tighter controls, and Amazon, Anthropic's largest investor. With Google — itself an Anthropic backer — signing anyway, the split leaves the safety-focused lab increasingly isolated on the year's central AI-policy fight.
The European Commission fined Google 890 million euros — about $1 billion — for breaking the Digital Markets Act, its first penalty against the company under the landmark competition law. Regulators split the fine in two: one part for ranking Google's own services above rivals in search, the other for blocking app developers from steering Google Play users toward cheaper offers elsewhere. Google has 60 days to change the practices or face further penalties, and says it is in "constructive" talks with Brussels.
Alphabet reported $119.8 billion in quarterly revenue, up 24%, with Google Cloud surging 82% to $24.8 billion and its cloud operating margin more than tripling to about 36%. Booming demand let the company frame its spending as justified: it raised full-year capital-expenditure guidance to between $195 billion and $205 billion and signaled a further jump in 2027. Gemini now reaches roughly 950 million monthly users.
Google committed $40 million in AI tools and cloud credits to the Genesis Mission, a White House initiative to double the pace of US scientific discovery within a decade. Tens of thousands of researchers across the US Energy Department's 17 national labs get a year of access to systems like AlphaFold, AlphaEvolve, and Gemini for Government. Early results are concrete rather than promissory: one materials lab cut microscope calibration from more than 90 minutes to about 13, freeing scientists to run real-time, autonomous experiments.
Google released Gemini 3.6 Flash, a faster, cheaper workhorse that posts large coding and agent gains while using about 17 percent fewer output tokens than its predecessor; Gemini 3.5 Flash-Lite, a high-throughput model running at 350 tokens per second; and Gemini 3.5 Flash Cyber, a security-tuned model that finds and fixes software vulnerabilities, offered in a limited pilot to governments and trusted partners. Notably absent was the flagship Gemini 3.5 Pro, which Google says is still in testing. The company also confirmed that pre-training has begun on Gemini 4.
Google is developing a server chip, internally dubbed Frozen v2, that hardwires parts of Gemini's architecture directly into silicon, The Information reported. The design could deliver six to ten times more tokens per unit of power than Google's current AI chips, easing the compute shortages that have limited Google Cloud. It is targeted for 2028 and meant to complement, not replace, Google's existing tensor processing units — another step in the industry's push to cut its reliance on Nvidia.
Google began rolling out Connected Apps for Search's AI Mode this week, letting US users link Instacart, Canva, and YouTube Music so the assistant can act, not just answer. Ask it to plan a barbecue and it scans your Google account and Gmail, assembles a cart, and hands it to Instacart for checkout; design and playlist requests route to Canva and YouTube Music the same way. Google frames it as the start of an "era of Search agents," though for now it teed up tasks rather than completing transactions itself.
San Francisco City Attorney David Chiu sent cease-and-desist letters ordering Apple and Google to stop distributing and profiting from 13 AI 'nudify' apps — marketed as face-swap tools but used to generate nonconsensual nude deepfakes of women and girls. Chiu argues the companies are 'aiding and abetting' the spread of the images and have likely made millions from in-app purchases tied to them. Google said it has already removed hundreds of such apps and takes swift action against violations; Apple did not comment before publication.
Three FireSat satellites — built by Muon Space for the nonprofit Earth Fire Alliance, with over 15 million dollars from Google.org and AI models from Google Research — launched on July 7th and are entering service just as thick wildfire smoke blankets the United States and Canada. Using multispectral imaging that can see through smoke, the constellation is designed to detect fires as small as five by five meters and flag them to agencies before they spread — early enough to matter, unlike most satellites that only see large, established blazes. The pilot has already spotted small fires other systems missed.
The European Commission issued two binding specification decisions under the Digital Markets Act. Rival AI assistants get the system-level Android access Google had reserved for Gemini — their own wake word, actions inside apps, and task delegation — starting July 2027. Separately, competing search engines and AI chatbots with search features can buy anonymized Google search data from January 2027. Google's Kent Walker said the orders introduce "unprecedented risks to user privacy, device security" and national security. Non-compliance under the act can cost up to 10 percent of global turnover.
NotebookLM becomes Gemini Notebook, and the rebrand carries a real capability with it: native code execution in a secure cloud sandbox, so the research tool can now run data analysis instead of only summarizing the sources you feed it. Notebooks also sync with the Gemini app, and Google says they will reach AI Mode in Search. It is live for AI Ultra and eligible Workspace customers now, and rolls out to Pro users on the web over the coming weeks.
Google Research introduced SensorFM, a foundation model pre-trained on more than one trillion minutes of sensor data from five million people wearing Fitbit and Pixel Watch devices. It draws on signals like heart rate, blood-oxygen, sleep, motion, and skin temperature, and transfers to 35 health-prediction tasks across cardiovascular, metabolic, sleep, and mental-health domains. When it grounded a personal health agent, its predictions matched the quality of using the real measurements.
When a fabricated image of Senator Mitch McConnell in a hospital bed spread across Reddit and X this week, fact-checkers at Snopes debunked it by detecting Google's SynthID watermark — an invisible signature baked into AI-generated images that survives screenshots, resizing, and compression. It was the watermarking system's first high-profile save. The catch: SynthID only works when the image generator participates, and while Gemini has embedded it since 2025 and OpenAI joined in May, most tools still don't.
Google released Gemma 4, a family of open-weight models from 2.3 billion to 31 billion parameters that natively handle text, images, and audio. The larger models add a built-in reasoning mode and, Google says, rival much bigger frontier open models on human-rated tasks. Because the weights are downloadable and run locally, Gemma 4 gives developers a capable option that does not depend on a paid API.
In a six-month study across 10 major US cities, Google tweaked Maps to steer a small share of drivers onto alternate routes with similar travel times, spreading traffic out instead of piling everyone onto the fastest road. The result was measurable: modestly higher speeds and lower fuel burn on targeted roads, adding up to thousands of tons of avoided carbon per city each year — a quiet example of AI optimizing a shared system rather than replacing anyone.
A new Google commercial reimagines the Founding Fathers writing the Declaration of Independence using Google Workspace — Docs suggestions, Meet calls, e-signatures, and Gemini taking notes — under the tagline "Group project, but make it 1776." Reception split sharply: praised on YouTube, panned elsewhere as "tone deaf," with historians arguing AI is the wrong metaphor for revolutionary collaboration. Critics also noted the ad itself appears partly AI-generated, feeding a broader debate over AI in creative and civic work.
Google DeepMind and the independent film studio A24 announced a research partnership that pairs the lab's researchers with A24's filmmakers to develop new creative tools and production workflows, with Google also taking an undisclosed equity stake in the studio. The companies named no specific models and set no fixed deliverables, framing it as open-ended, exploratory work. It is a notable move as frontier labs court Hollywood, arriving amid ongoing tension over AI's role in creative work.
Google's new environmental report shows its electricity use climbed 37 percent in 2025 — its largest single-year jump ever, and more than 250 percent above 2019 — driven mostly by AI. The honest counterpoint: Google still cut operational emissions 2 percent by matching all that power with renewable purchases, and it signed over 12 gigawatts of new clean energy last year. Its own conclusion is the sobering line — the AI buildout "is currently accelerating faster than the grid is decarbonizing."