Google DeepMind has partnered with the U.K. government’s AI Security Institute (AISI) to turn “cutting-edge AI into real benefits,” according to a press release published Thursday. DeepMind will open its first automated research lab in the U.K. next year. The lab will be fully integrated with Gemini and will focus on developing new superconductor materials that can carry electricity with zero resistance.
Technology Secretary Liz Kendall signs the government’s new partnership with Google DeepMind
DeepMind for Science
This agreement expands Google DeepMind’s collaboration with the U.K government to “speed up progress in science and education, modernize public services and advance national security and resilience,” according to a press release from DeepMind published Wednesday.
The automated lab will be able to synthesize and characterize hundreds of materials per day, shortening the timeline for identifying new materials that could “tackle critical energy challenges by unlocking advanced batteries, next-generation solar cells and more efficient computer chips,” a press release stated.
DeepMind will also work with the government and U.K. scientists to conduct AI-enabled research to advance cheaper and greener energy, such as fusion power. The press release also stated that a “potential” Gemini for Government system could automate routine tasks for civil servants.
Under the agreement, U.K. scientists will be given priority access to DeepMind’s AI models and tools, including AlphaGenome, AlphaEvolve, WeatherNext and AI co-scientist.
“This agreement could help to unlock cleaner energy, smarter public services, and new opportunities which will benefit communities up and down the country,” said Technology Secretary Liz Kendall.
DeepMind for public services and security
The U.K. government’s AI Incubator team is currently trialing Extract, a tool for council planners that uses Gemini to convert old planning documents into digital data in just 40 seconds per document.
DeepMind will work with the U.K. government to advance national cyber resilience with AI through tools such as Big Sleep (for unknown security vulnerabilities) and CodeMender (creates patches and rewrites codebases)—a proactive coding assistant that can rewrite existing codebases to eliminate vulnerabilities—to identify issues and automatically fix code.
DeepMind will also focus on developing a version of Gemini tailored for teachers, founded in the national curriculum, and tested for safe use in schools. The company will work with the U.K.’s Security Institute on foundational research to ensure AI is safe and beneficial.
DeepMind will build on previous research into techniques to monitor an AI system’s chain of thought (CoT) and into “socioaffective misalignment”—the risk that AI models may behave in ways that harm users, such as manipulation or deception, even when the model is following its programming instructions.
