News, updates, and in-depth coverage
Google conducts artificial intelligence research and development primarily through Google Research and Google DeepMind, its dedicated AI research division formed in 2023 through the consolidation of Google Brain and DeepMind. Google develops large-scale machine learning systems used across its products and services, including search, advertising, cloud computing, and consumer applications. Its AI research spans areas such as large language models, multimodal systems, reinforcement learning, computer vision, robotics, and scientific discovery. Google DeepMind is responsible for foundation model development, including the Gemini family of multimodal models, as well as research in alignment, reasoning systems, and applied AI in fields such as biology and materials science.
Google Research announces NAI, a framework for building AI that adapts to users with disabilities from the start rather than bolting on accessibility later. AI agents can reconfigure themselves to generate audio descriptions for blind users or simplify layouts for ADHD. Google.org is funding RIT/NTID, The Arc, RNID, and Team Gleason to build adaptive tools, including Grammar Lab - a Gemini-powered tutor for deaf students.
Google disclosed Gemini now has over 750 million MAUs during Q4 2025 earnings—up from 650M last quarter—though still trailing ChatGPT's estimated 810M users.
Alphabet announced massive 2026 capital spending as AI demand outpaces compute. Gemini now processes over 10 billion tokens per minute via APIs, up from 7B last quarter.
Google partners with the Vertebrate Genomes Project and Earth BioGenome Project to sequence genomes of endangered species including cotton-top tamarins, ibex, and penguins. AI tools like DeepPolisher, DeepVariant and DeepConsensus have dramatically reduced sequencing time and cost. Google.org is funding Rockefeller University to expand to 150 additional species, all openly released for conservation research.
Google DeepMind releases Project Genie to AI Ultra subscribers in the US - an experimental prototype powered by Genie 3 that generates explorable, interactive worlds in real time. Users can create environments via text/image prompts, explore as the world generates ahead of them, and remix existing worlds. Unlike static 3D snapshots, Genie 3 simulates physics and interactions dynamically. Current limitations include 60-second generations and occasional physics/control issues.
Google added agentic vision to Gemini 3 Flash, combining visual reasoning with code execution. Rather than single-pass image analysis, the model now investigates images step-by-step, manipulating and verifying details.
Google AI Plus expands to 35 new countries including the US, offering Gemini 3 Pro, Nano Banana Pro, AI filmmaking in Flow, and NotebookLM research tools. Includes 200GB storage and family sharing for up to 5 members. Existing Google One Premium 2TB subscribers get automatic access. Launch promo: 50% off first 2 months.
No articles yet