Google is rolling out new model updates for Gemini Live on Android and iOS, giving its voice assistant more natural speech, flexible pacing and playful accents so it can help with studying, language practice, role play and storytelling.
Google is adding Gemini-powered audio overviews to Drive, turning long PDFs into short spoken recaps you can play from your files. It is fast, tidy, and built for days when reading has to wait.
Google has announced a new AI processing platform called Private AI Compute that aims to deliver the same level of privacy and security as on-device processing while using advanced cloud models.
A new class of adaptive malware called PROMPTFLUX — capable of mutating its own code via LLMs - represents a major escalation in cyber-threats, according to Google’s GTIG.
Google is rolling out Gemini on Android Auto for some users today. The assistant replaces Google Assistant, keeps the "Hey Google" hotword, adds Live for back and forth, and drops contact nicknames. Message translation is included.
Apple may be turning to Google’s Gemini AI to power a smarter version of Siri. Reports suggest the upcoming Apple Intelligence platform could use a custom Gemini model, boasting 1.2 trillion parameters, to handle complex reasoning and multi-step tasks while Apple continues developing its own in-house AI systems.
A new Penn State study shows you don’t need hacking skills or prompt-engineering tricks to break past AI safety rules. Ordinary users armed with nothing more than natural conversation were able to trigger biased and stereotype-driven responses across major chatbots like ChatGPT and Gemini. The finding flips the script on AI safety, suggesting real-world conversations are just as important to test as lab-style attacks.
Early access for Gemini for Home is rolling out to US users, and it works on every Google speaker and display since 2016. Expect more natural voice control now, with broader locales planned in early 2026.
Google’s Gemini for Home system, recently integrated with Nest Cams and other devices, is already drawing attention. Though, not for its slick smart-home control but for mis-identifications like mistaking a dog for a deer and detecting fake people. The mishaps underscore both the promise and early teething problems of AI in home security.
Google Gemini has evolved into a powerful AI platform with deeper integrations, smarter models, and new capabilities. Here’s everything you need to know about how it works in 2026.
Google is rolling out Gemini for Home in early access, bringing its AI smarts to Nest speakers and displays. The new update lets you have more natural conversations, control smart devices with detailed voice commands, and even search camera history using simple phrases. With smarter routines and richer context, your home assistant just got a serious upgrade.
Google’s Gemini in Canvas can now turn your ideas, notes, or files into full presentations in seconds. The AI generates slides complete with text, layouts, and images, all ready to export to Google Slides. It's your new shortcut from blank page to polished deck.
A BBC and EBU study found that nearly three out of four of Google Gemini’s news answers contained major errors or sourcing issues, making it the least reliable AI assistant for current affairs. Researchers say even the best AI tools still struggle to separate fact from fiction.
Samsung used its Galaxy XR reveal to confirm AI glasses are on the roadmap, tied to Android XR and Gemini. Fashion partners hint at frames you’ll actually wear, not another lab demo.
Chrome’s Canary build quietly adds a “Contextual tasks” panel tied to Gemini. It's rough, but it points to agentic browsing, useful in small steps, as long as permissions, controls, and trust catch up.
OnePlus 15 will be the first phone to run OxygenOS 16, which is built atop Android 16 and brings a whole bunch of AI-powered tricks to phones to and tablets.
AI use in research has surged, but rising concerns about accuracy, ethics, and transparency mean institutions must step up with better tools, training, and disclosure standards to unlock AI’s full potential.
Researchers have revived a 12-year-old data theft technique, now called Pixnapping, that lets malicious Android apps steal sensitive on-screen data — including 2FA codes — without permissions or user knowledge.
NotebookLM’s Video Overviews just got Nano Banana visuals and a new Brief format, giving visual learners faster, clearer summaries with styles you can tune on the fly.