Google AI Updates: September 2, 2026

1. Gemini Gets Agentic Video Understanding

Google DeepMind. Gemini 3.7 Flash, 3.6 Flash, and 3.5 Flash-Lite can now decide which segments of a video to examine and at what resolution, instead of sampling every frame at a fixed rate. Google reports token usage falling by up to 88% while accuracy improves, with the biggest gains on multi-hour footage where uniform sampling wastes most of the budget on irrelevant frames. The change turns video analysis into a search problem the model drives rather than a fixed preprocessing step. Source

2. Google Pics Brings Nano Banana Image Editing Into Workspace

Google. Google Pics is now available in Google Workspace as an image creation and editing tool built on the latest Nano Banana model, with prompting rather than a canvas as the primary interface. It targets the same jobs Canva and Adobe Express serve, but the workflow is describing the edit instead of manipulating layers. Availability is tied to Workspace rather than a standalone consumer app. Source

3. Android Update Wires Gemini Into Accessibility Features

Google. The latest Android release adds motion sickness mitigation and a set of accessibility improvements, several of which use Gemini to generate or adapt content on device. Some features close gaps with equivalents Apple already ships on iPhone, while others depend on the model being present locally. The update ships through the standard Android feature drop channel rather than a platform version bump. Source