Google AI Updates: May 30, 2026

1. Google Details Gemini Omni Video Model and Gemini 3.5 Flash

Google. Google published a set of demos for Gemini Omni, a video generation and editing model that takes images, audio, video, and text as input and supports conversational, multi-turn edits while holding character consistency and scene physics across changes. Alongside it, Google positioned Gemini 3.5 as a family pairing frontier reasoning with action, launching first with 3.5 Flash for agentic workflows and coding. Both are rolling out across the Gemini app, AI Studio, the APIs, Search, and Workspace, with Omni available to AI Plus, Pro, and Ultra subscribers globally and free access arriving via YouTube Shorts and the YouTube Create app. The combination shows Google folding generative video and a faster agentic model into the same release cadence, with distribution through its own consumer surfaces as the wedge. Source

2. Google Cloud Launches AI Threat Defense for Automated Vulnerability Remediation

Google Cloud. Google Cloud introduced AI Threat Defense, a security platform that combines Gemini, Wiz, and DeepMind technology to scan enterprise networks, simulate exploits, and deploy automated code fixes for vulnerabilities. It targets AI-accelerated attacks that can weaponize newly disclosed flaws within hours, positioning autonomous remediation as a replacement for manual patch cycles that cannot keep pace. For security teams the pitch is collapsing the window between disclosure and fix from days to minutes, though handing automated code changes to a remediation agent raises its own review and trust questions. Source

3. Google Unveils the Coral Board, Running Gemma 3 Fully Offline

Google. Google launched the Coral Board, a compact single-board computer built on a Synaptics chip with 2GB of RAM and roughly 1 TOPS that runs the Gemma 3 270M model entirely on-device with no cloud connectivity. Demos covered real-time translation, voice-controlled hardware, and generative music, with the supporting code released as open source on GitHub. It gives developers a concrete reference platform for privacy-preserving, on-device small-model deployment at the edge, an area where most options are still hobbyist kits rather than first-party hardware. Source