Google Rolls Out Genie 4 World Model and Project Genie Agent to AI Ultra Subscribers
Google has released Genie 4, its most realistic interactive world model, to Google AI Ultra subscribers. It supports mid-session prompting to edit environments in real time and adds native real-time audio. The update also brings Genie 3 upgrades, a Project Genie agent that coordinates Google's multimodal models, and a community gallery for sharing and remixing worlds, with SynthID watermarking and safety screening applied.
Skynet Chance (+0.01 pp): World models are a consumer creative tool here, with watermarking and safety screening, so the direct control risk is minimal. They do incrementally advance AI's ability to model and simulate environments, which agentic systems could later use.
Skynet Date (no change): More capable world simulation slightly speeds the development of agents that can plan in rich environments. The effect on the risk timeline is modest.
AGI Progress (+0.02 pp): Real-time promptable world models with improved physics, audio and consistency advance learned world simulation, which is widely seen as key to general agents and embodied AI. Google acknowledges that limitations in long-term scene permanence and physical interactions remain.
AGI Date (no change): Rapid iteration from Genie 3 to Genie 4 within a year, plus wider deployment, suggests world-model capabilities are maturing quickly. This modestly accelerates the timeline toward training environments for general agents.
<< All AI news for November 18, 2026
[ Get the daily index digest on Telegram → ]Related AI News
- DeepMind's AlphaProof Code Formally Verifies Key Signal Protocol Guarantees in Four Weeks
- DeepMind's AlphaProof Code Formally Verifies Signal Protocol Implementation
- Google DeepMind Launches Gemini Robotics 2.5 for General Real-World Dexterity
- Israeli Startup 42AI Emerges from Stealth with $62.5M for Behavior-Predicting World Model
- Google DeepMind Unveils Gemini 4 Carbon with Leading Cybersecurity and Multi-Day Agentic Capabilities