Gemini Robotics 2.5: DeepMind's New Robot Model
DeepMind says Gemini Robotics 2.5 lifts robot task success 30% to 100% and hit 96% in a Mercedes-Benz kitting pilot. It is rolling out to partners only.

News on DeepMind, Google's AI research lab. Coverage includes DeepMind's double-blind AI evaluations to prevent benchmark contamination, the Backstory tool for image fact-checking, and Google folding DeepMind into the main business.
DeepMind says Gemini Robotics 2.5 lifts robot task success 30% to 100% and hit 96% in a Mercedes-Benz kitting pilot. It is rolling out to partners only.

DeepMind says AlphaProof Code proved key security properties of Signal's libsignal encryption code in about four weeks, not the whole Signal app.

Google DeepMind's AlphaGenome 2 predicts how DNA variants affect specific human cell types. It is in the API for non-commercial research now.

Human-guided AI agents at DeepMind found cheaper attacks on legacy 1024-bit RSA. What ran, what is only estimated, and what it means for 2048-bit keys.

Google DeepMind and Fab AI will back local teams building teacher tools on open Gemma models in Kenya, Ghana and South Africa. What Google has said so far.

AlphaFold is Google DeepMind's AI for predicting protein structures. What it does, AlphaFold 2 vs 3, how to use it free, confidence scores and its limits.

SynthID is Google DeepMind's invisible watermark for AI-made images, video, audio and text. Here is how it works, how to check a file in Gemini, what each result means, and where the watermark stops working.

Why AI labs built ever larger models, and why the 2022 Chinchilla paper changed course: a plain-English guide to scaling laws, how to split a compute budget between model size and data, and the newer research on scaling the time a model spends answering.

Google DeepMind researchers argue smarter AI shifts wages gradually, while AI agents that coordinate well could abruptly reshape firms, jobs and markets.

Reactions to DeepMind's Genie 4: praise on X for realism and live steering, physics errors spotted on Hacker News, and complaints about the waitlist.

DeepMind's Genie 4 world model adds native audio and dialogue, more lifelike people and physics, and live text edits. Trusted testers only for now.

Gemini Robotics ER 2, Google DeepMind's robot planning model, is in public preview on Google Cloud. What it does, Google's numbers and its $1 to $10 token prices.

The DeepMind Institute publishes essays by Google DeepMind and Google researchers on safe AGI, jobs and society. Here is who runs it and what it has published.

The UK plans to legislate against non-compete clauses, with details due at the October 28 Budget. Here is what was announced, what is still open, and why it matters for DeepMind's 6 to 12 month restrictions and its talent losses to Anthropic and OpenAI.

Google DeepMind researchers have released empirical results detailing an expanded version of Co-Scientist, transitioning the Gemini-based multi-agent system from purely computational hypothesis generation into a closed-loop research framework connected to physical laboratory instruments and automated code pipelines. Documented in an arXiv preprint by Samuel Schmidgall, Zhu, and collaborators, the system combines multi-agent ideation, deterministic execution scaffolding, and automated manuscript

Google DeepMind, in collaboration with the Singapore AI Safety Institute, OpenMined, AVERI, and MLCommons, has piloted a cryptographic framework for double-blind evaluations of proprietary frontier language models. The pilot, conducted on Gemini 2.5 Flash Lite, uses hardware-isolated confidential computing to ensure that model developers cannot see evaluation prompts while evaluators cannot inspect proprietary weights or inference code. The project addresses benchmark contamination and intellec

Google DeepMind has detailed its 15-year trajectory of game-based artificial intelligence research, outlining how milestones from arcade reinforcement learning to modern multimodal models have culminated in an experimental research program inside the persistent virtual universe of EVE Online. The retrospective connects early breakthroughs in discrete, fully observable games to the frontier challenges currently facing autonomous systems: long-horizon planning, non-stationary multi-agent dynamics

Nonprofit AI safety evaluation organization Guidelight has published its inaugural assessment of internal control practices across frontier AI developers. Evaluating public disclosures from Anthropic, OpenAI, Google DeepMind, xAI, and Meta, the study finds that foundational mechanisms for monitoring, gating, and containing advanced internal AI models remain only partially implemented across the industry. No evaluated organization achieved full or near-full implementation on any of the standard'

Google DeepMind has expanded live testing of Backstory, an experimental verification platform designed to investigate the origin, manipulation, and dissemination history of digital images. The system, built on the Gemini model family, is currently deployed across newsrooms, open-source intelligence (OSINT) groups, academic researchers, and fact-checking teams participating in Google's Trusted Testers program. Beyond Binary Synthetic Detection Traditional automated image forensic tools typical

Google DeepMind has introduced WeatherNext Cyclones, or WN-C, an AI system for tropical cyclone forecasting that sees about one day further into the future than leading operational models. The gain roughly matches the progress traditional weather models made over a decade. The model was built with the National Hurricane Center, the Cooperative Institute for Research in the Atmosphere, and the UK Met Office. Forecasts have been running live on Google's Weather Lab since June 2025. During Hurrica
