Google has outlined a series of artificial intelligence product launches and updates across its consumer, developer and research businesses. The changes include Gemini 3.5 Live Translate and new features in Android 17.
The update spans software for phones and laptops, tools for app developers, education products, research systems and a new Google Home Speaker built around Gemini.
Consumer tools
Among the most prominent additions is Gemini 3.5 Live Translate, an audio model for speech-to-speech translation. The system can automatically detect more than 70 languages, preserve a speaker's intonation and reduce pauses in live conversations.
The feature is rolling out through the Gemini Live API, Google AI Studio and the Google Translate app. It is aimed at multilingual calls, meetings and travel.
Android 17 is also central to the latest product push. The mobile operating system adds floating app windows, picture-in-picture recording reactions, changes for foldable gaming layouts and new security measures, including the option to lock a missing phone using biometrics.
Those changes are arriving first on Pixel devices before wider distribution to eligible Android devices. Separate updates in the latest Pixel Drop include voice translation changes, custom voicemail greetings and automated emergency notifications.
Google is also bringing its Finance service out of beta. Users will be able to track personal portfolios, follow market developments and use an Android app with an AI research tool and AI-generated explanations of stock price moves.
Another hardware addition is a new Google Home Speaker designed around Gemini. The device is intended to support more natural spoken interactions, handle several requests in a single exchange and maintain context during conversations.
Developer push
For developers, Google highlighted Gemma 4 12B, a local model that runs on a laptop with 16GB of memory. The model combines vision and voice processing in one system, allowing users to run AI tasks on their own hardware rather than relying entirely on cloud services.
Google has also integrated computer use into Gemini 3.5 Flash. That change allows developers to create agents that can work across desktop, mobile and browser environments for tasks such as software testing and office-based workflows.
Two further model updates target image and video creation. Nano Banana 2 Lite is now available as Google's fastest and lowest-cost Gemini Image model, while Gemini Omni Flash has entered public preview through APIs as a multimodal model for video workflows.
Education focus
Education featured heavily in the update. NotebookLM has been upgraded with stronger reasoning, a cloud computer for running code, and tools to generate charts, spreadsheets and slide decks.
The service can also help users organise ideas and collect web sources into a structured research repository. Access is being provided globally to Google AI Ultra subscribers and some Workspace accounts.
The Gemini app is also receiving study notebooks. Students can set a learning goal, upload notes, take a baseline quiz and receive lessons tailored to their needs, with progress tracked through a dashboard.
Schools are another target. Updates across Google Classroom, Chromebooks and Gemini are intended to give teachers more help with administrative tasks and curriculum-linked activities, while students gain access to study notebooks and test preparation tools.
Google also pointed to research on AI in education, including a study in Sierra Leone focused on classrooms where teacher shortages are acute. It has also released a teacher training guide and research playbook linked to that work.
Research and public services
In science, Google said research teams are using its Co-Scientist system to develop and refine hypotheses in life sciences. It cited work on infectious diseases, cellular ageing and ALS as examples of current use.
Public services are another area of focus. A Gemini-based planning prototype developed with Google DeepMind, Faculty and the UK government is designed to address administrative backlogs and case analysis in local councils.
Google said the experiment could reduce household planning application workloads by 50% through automation of data extraction and policy cross-referencing.
Google also disclosed legal action against what it described as an organised cybercrime network known as the Outsider Enterprise. The group distributes phishing kits through Telegram for fake text campaigns that mimic trusted brands, including Google, it said.
Alongside that action, Google said it is backing seven bipartisan bills aimed at tackling scams, including those using AI, while deploying AI-based systems to detect such threats.
Climate and culture
Google said it has updated models used for flood, wildfire and cyclone forecasting. The systems are now being used to predict river floods seven days ahead, track wildfire boundaries from satellite imagery and forecast cyclone paths, with alerts distributed through Search and Maps.
The company also used the update to highlight cultural projects. One is a collaboration with Colonial Williamsburg that includes a digital collection and a custom NotebookLM drawing on more than 150 primary sources and articles.
Another is Dataland, which Google described as the world's first museum dedicated entirely to AI arts, created with media artist Refik Anadol.
On workplace adoption, Google cited research conducted with Public First on AI use in the UK. The study found that workplace adoption had risen to 73% from 34% in 2025.
"The top 15% of UK AI users are benefitting from the technology. They are fast-tracking their careers and more likely to report strong performance reviews, promotions and pay rises," Google said.