Body
Google used its annual developer conference, Google I/O, to preview a new wave of AI features it said will push Gemini beyond conversation and into task execution across Google’s apps. Speaking Tuesday near the company’s Mountain View, California headquarters, CEO Sundar Pichai described the company’s focus as “agentic” AI and framed the next step as making agents more useful while improving security.
Pichai said in the keynote that Google was “firmly in our agentic Gemini era,” and he added that while he had experimented with agents and saw potential, they were still at an early stage for making them easy to use, “super secure and truly helpful.” Google also tied the rollout to its broader AI spending, saying on an investor call in late April that this year’s capital expenditures could rise as high as $190 billion, while pointing to strong quarterly earnings growth and a reported stock gain of about 11% since the prior month’s report.
In a separate set of comments during the keynote, Google said its Gemini app had 400 million monthly active users last year and that usage has since surpassed 900 million, more than doubling in a year. The company said those user numbers accompany a major update to its model lineup that it plans to roll out in stages.
One of the updates is Gemini 3.5 Flash, which Google said is beginning Tuesday as a default model for the Gemini app and for “AI mode” on Google search. Google described Gemini 3.5 Flash as focused on speed, saying it is its strongest agentic and coding model yet and that it is about four times faster than some competitors, while also describing the model as having been developed with new safety training and mitigations designed to reduce the chance of harmful outputs or incorrect refusals to answer safe queries.
Google also previewed Omni, a new model it said will let users create high-quality video by entering a query with various kinds of input, including text, images, videos and audio. Google said users will eventually be able to create images and audio with Omni, but it did not provide specific timing for those features. For the video component, Google said Omni can produce results that are more realistic than videos made by other models because of what it described as the model’s understanding of physical forces such as gravity, kinetic energy and fluid dynamics.
As part of its effort to pair generation with verification, Google said Omni’s videos will include the company’s imperceptible digital watermark, SynthID. The company also said it will add “content credentials verification” to the Gemini app—describing a tool that determines whether content such as photos or videos was created by AI or captured with a phone camera and then edited with AI tools—and said that feature will appear in search in Chrome in the coming months. Google also said OpenAI, Kakao and Eleven Labs have adopted SynthID technology for more of their AI-generated content.
Alongside models for media creation and search, Google introduced Gemini Spark as an agent that it described as operating continuously. Powered by Gemini 3.5, Spark is described as able to sort through routine work such as meeting notes, emails and chats, and then produce a document summarizing takeaways and to-dos. Google also said Spark differs from other available agents because it is based in the cloud, so it can keep working in the background even when users shut laptops or lock phones.
Google said Spark’s proactive behavior is also why the company is emphasizing permissions and risk controls. The company said Spark is designed to ask permission before performing “high-stakes” tasks like sending an email or making a purchase. Google said select testers will have access beginning Tuesday and that it plans to roll out a beta mode to U.S.-based subscribers in its Google AI Ultra tier, with a further expansion planned for later this summer to run directly within Chrome.
Google also used the conference to provide more detail on smart glasses it has been developing. The company said it would offer two kinds: audio glasses that provide spoken help in the ear and display glasses that present information visually. Google said the audio glasses are expected to arrive later this fall, and it said users will be able to say “Hey Google” or tap the side of the frame to access Gemini, which would then assist with functions including navigation, managing phone communications and real-time translations.
Google said it is partnering with Samsung, as well as eyewear brands Gentle Monster and Warby Parker, and it showed first looks at two designs on Tuesday. The company said the sunglasses from Gentle Monster and the glasses from Warby Parker will launch as part of those brands’ full collections later this year.
In search and shopping, Google said it is building further on the “AI mode” feature introduced last year, which it described as giving users a more conversational response before providing links. Google said AI mode queries have “more than doubled every quarter” since its launch and that the tool recently surpassed 1 billion monthly users, according to Liz Reid, Google’s head of search. Google also said Gemini 3.5 Flash will become the default model in search and that it will introduce an “intelligent search box,” described as adapting to longer queries and helping users write questions with AI-powered suggestions rather than only traditional autocomplete.
Google said it is also expanding how people can search by using multiple modalities, including text, images, video, files and even Chrome tabs as inputs. It said the new search box rollout begins Tuesday in countries and languages where AI mode is already available, and it described the next step in shopping as the Universal Cart, which it said would work across merchants and services and run on Gemini models to look for deals, price drops, price history information and restock alerts. Google said Universal Cart will be available on search and the Gemini app this summer, with YouTube and Gmail to follow.