Google's Gemini hits 900m users

Google said on May 19 that Gemini had reached 900 million regular users, a figure the company disclosed at its I/O developer conference as it widened Gemini’s role across search, Android and creative tools. The update came as Britain’s AI Security Institute said it had built a government research team with more than 100 technical staff, including alumni from OpenAI and Google DeepMind. The two developments point to a race now being fought on two tracks: user distribution inside large consumer platforms and technical state capacity around model safety. Google and the British institute each described the shift in operational terms — users, staff, compute, access and testing. ### Where did the 900 million figure come from? Google CEO Sundar Pichai gave the number in his opening keynote at Google I/O 2026 on May 19, saying the company was measuring “people regularly using” Gemini across its products. The Indian Express, citing Google’s announcement, reported that the total had more than doubled in a year and was roughly in line with OpenAI’s self-reported active-user scale. Pichai used broader usage data to frame the number. Google said it was processing more than 3.2 quadrillion tokens per month across its surfaces, that more than 8.5 million developers were building with its models monthly, and that about 375 Google Cloud customers had each processed more than one trillion tokens over the past 12 months. (indianexpress.com) ### Why does Google’s distribution matter here? Google entered the generative AI race with products that were already embedded in Search, Android, Chrome, Workspace and YouTube. The Indian Express said that installed base is helping Gemini close ground because Google can place AI features inside products that already reach billions of users, rather than relying only on a standalone chatbot app. (blog.google) Google’s own I/O materials showed the same approach. The company said Gemini models were helping drive usage across its products and noted that it now has 13 products with more than 1 billion users each, including five with more than 3 billion users. ### What is Gemini Omni Flash, exactly? Google introduced Gemini Omni Flash on May 19 as the first model in its Omni family, describing it as a system that combines Gemini’s reasoning with generative media models and starts with video. (indianexpress.com) The company said the model is available in the Gemini app, Google Flow and YouTube Shorts. (blog.google) Google said Omni Flash can take text, image, audio and video as input to generate video outputs and can edit video through conversational prompts. In its model card, Google DeepMind said the release followed red teaming and evaluations conducted with internal safety, security and responsibility teams. (blog.google) ### What is Britain building on the safety side? Britain’s AI Security Institute said it is a directorate of the Department for Science, Innovation and Technology focused on research that gives governments a scientific understanding of advanced AI risks. On its public site, the institute said it has more than 100 technical staff and includes senior alumni from OpenAI, Google DeepMind and the University of Oxford. (blog.google) The institute said its work includes testing leading AI systems before public release, informing policymakers in Britain and allied governments, and funding research into mitigations. Its leadership includes Interim Director Adam Beaumont, a former GCHQ chief AI officer; Chief Technology Officer Jade Leung, previously head of governance at OpenAI; and Chief Scientist Geoffrey Irving, who has led teams at OpenAI and DeepMind. (aisi.gov.uk) ### What does the institute say it is doing with that capacity? The AI Security Institute said its research agenda is centered on understanding advanced model capabilities, building assessment infrastructure and developing risk mitigations. The institute described itself as the world’s largest government team dedicated to AI safety and security research and said it works with the National Cyber Security Centre, the Defence Science and Technology Laboratory and frontier AI companies. (aisi.gov.uk) AISI has also tied that work to grants and external partnerships. In February it said the first cohort of its Alignment Project had taken total coalition funding to 27 million pounds, and its website says applications are expected to reopen in summer 2026. ### What should readers watch next? May 19 remains the key date for Google’s latest push, because that is when the company launched Gemini Omni Flash and published the accompanying model materials. (aisi.gov.uk) Google said additional output modalities, including image and audio, would come later. Summer 2026 is the next dated milestone on the British side. (aisi.gov.uk) The AI Security Institute says the Alignment Project is expected to reopen then, while the institute continues publishing research on model oversight, cyber risks and evaluation methods. (aisi.gov.uk) (blog.google)