Google DeepMind — Latest News and Analysis

Google’s new Genie 4 AI world model research preview generates interactive worlds with sound and responsive dialogue. Trusted testers can try it out to reshape scenes with text instructions.

Four Google Gemini 4 Argon AI agents were able to defeat Minecraft's final boss Ender Dragon, writing 2,783 programs and sharing information between each other.

Google’s public SynthID checker looks for supported AI watermarks in images, videos and audio. A missing mark leaves a file’s origin unresolved.

Eight former OpenAI, Anthropic and Google DeepMind employees describe different reasons for leaving, including limits on speaking publicly and choosing their research.

Anthropic proposes ongoing access for outside reviewers to inspect AI development and publish findings, making its promises checkable; slower capability growth across labs requires coordination.

OpenAI's ChatGPT Images 2.5 model preserves more original details when users revise an image’s setting, style or composition.

In a Google DeepMind experiment, AI math agents copied accepted shortcuts, closed unsolved tasks and left peers able to expose cheating but unable to reverse it.

Google DeepMind's AlphaGenome Atlas makes predictions for roughly nine billion DNA substitutions searchable for non-commercial research and experimental triage.

WeatherNext 3 brings hourly weather forecasts, 15-day main cycles and finer near-surface detail, with data available through Google Cloud and Earth Engine.

A team of eight researchers reports that encrypted reasoning blocks returned by major AI model APIs can sometimes be recovered.

Google has released lower-cost Gemini 3.6 Flash and Flash-Lite models and plans a restricted Gemini 3.5 Flash Cyber pilot for cybersecurity testing.

Google has reportedly delayed Gemini 3.5 Pro after missing a June target as coding results fell short, leaving partner testing underway without a release date.

Google DeepMind CEO Demis Hassabis proposes a US-led body to test frontier AI before launch and potentially gate US deployment.

Google has released Nano Banana 2 Lite and Gemini Omni Flash in preview, pairing lower-cost image generation with short-video editing for developers.

Google DeepMind and A24 plan AI filmmaking tools with creator feedback and library-data limits, testing whether storyboards aid artists without catalog access.

Nobel prize winner and AlphaFold pioneer John Jumper is leaving Google DeepMind for Anthropic, adding scientific weight to the talent fight among labs.

Google DeepMind has shared its AI Control Roadmap for agentic access controls and monitoring as it prepares for alignment failures and wider agentic tool access.

Google has introduced DiffusionGemma to speed local AI output through parallel text diffusion, but lower quality than Gemma 4 keeps trade-offs visible.

AI leaders have backed DNA and RNA screening rules that would make gene synthesis sellers verify customers and orders before risky designs reach labs.

Sakana AI has opened a Recursive Self-Improvement Lab to test whether AI can cut compute dependence.

Google's Gemma 4 QAT models include a sub-1 GB E2B text-only setup for lower-memory local AI on laptops, phones, GPUs, and edge devices.

Google has started a pilot which pays Play Store developers for private Android app code, leaving pricing, retention and AI-training limits unclear.

Google has released Gemma 4 12B, a local multimodal AI model for laptops that tests whether audio, images, code, and tool calls fit in 16GB memory locally.

DeepMind’s AlphaProof Nexus solves nine Erdős problems using Lean-checked proofs, signaling a new phase in AI math after OpenAI’s geometry claim.

Google has added Street View grounding to Project Genie, letting AI Ultra users start simulated worlds directly from real U.S. locations with Maps imagery.

A possible Google DeepMind deal with Contextual AI would bring over 20 researchers and a software license, extending its Hume-style acquihire playbook.

Google's wider CodeMender test would give select experts new API access to its AI code-security tool while keeping human review on every proposed patch.

Google has launched Gemini Omni with Omni Flash for paid users first, bringing multimodal video generation to its apps before wider API access arrives.

Demis Hassabis' possible early Anthropic stake has added a personal layer to Google's cloud and TPU ties as Claude and Gemini compete for enterprise AI buyers.

Stanford HAI has launched the AI and Organizations Lab to study how workplace AI changes jobs, teams, decision-making, and accountability with research.