Wednesday, 11 November 2026
DeepMind's AI Agents Prove Signal's Real Code Keeps Its Security Promises
DeepMind’s AI Agents Prove Signal’s Real Code Keeps Its Security Promises
Google DeepMind says it has formally verified the actual code behind the Signal protocol — the encryption that protects messages in Signal and Google Messages — and not just its design on paper. The method, called AlphaProof Code, splits the work: human experts decide what must be true, researchers turn those security goals into formal specifications that people review one by one, and agents running on Gemini Carbon write the proofs in Lean, a language a machine can check line by line. The agents produced more than six million lines of Lean in four weeks and flagged about one percent of the codebase for a human second look; Lean checked the rest. As a test, the team put two real, long-fixed Signal bugs back into an isolated copy of the code, and both times the system proved the buggy version insecure. DeepMind calls it the first time the implementation of a widely deployed protocol like this has been formally verified by any method, and is open-sourcing the proofs. Work it says used to take expert teams years — the question now is whose code gets proved next.
Gemini 4 Carbon Opens to Paying Users — and Its Price Doubles After November 30
Two weeks after Google gave Gemini 4 Carbon to cyber defenders first, the model is now open to every paying developer and business through Google AI Studio and Gemini for business, and is rolling out to Google AI Ultra subscribers in the Gemini app, AI Mode in Search and Antigravity. The new part is the bill: Carbon costs $2 per million input tokens and $10 per million output tokens until November 30, the same as Gemini 4 Argon — then $4 and $20. Google’s own figures put it first on CursorBench 4.0, Perplexity’s WANDR and Zapier’s AutomationBench at less than half the cost per task of the next-best outside model. In simulated cyberattacks it blocked 99% of attack paths, and its per-attempt breach rate fell to 0.8% from Argon’s 3.1%. All of these are the maker’s measurements. An introductory price with an end date — a discount, or a countdown?
Source: Google, November 10, 2026
Humanoid Robots Running Gemini Robotics 2.5 Sort Parts on Mercedes-Benz Lines
Google DeepMind released Gemini Robotics 2.5, its new model for controlling robot arms and hands, and showed it at work outside the lab. Apptronik’s Apollo 2 humanoids ran it in a proof of concept on active Mercedes-Benz production lines in Kecskemét, Hungary, and Sindelfingen, Germany, picking parts such as metal brackets out of bulk bins and laying them into trays in the order each car needs them — with a 96% success rate. A second model, Gemini Robotics ER, keeps track of the workflow while 2.5 moves the hands. Across 200 evaluation tasks, DeepMind reports 30% to 100% higher success than its previous models and work done 20% to 30% faster. Extra reasoning at run time raised safe task completion near people by 43%. Not a dancing demo this time — a tray of brackets, done in the right order.