Google DeepMind has expanded its partnership with the UK AI Security Institute (AISI) through a new Memorandum of Understanding focused on foundational security and safety research. This collaboration aims to ensure AI development benefits humanity while prioritizing safety and responsibility. The partnership builds on existing work since AISI's inception in November 2023 when they began testing Google's most capable AI models.

Under the expanded agreement, both organizations will share proprietary models, data, and ideas to accelerate research progress. Teams will collaborate on joint reports and publications while combining expertise to tackle complex safety challenges. Key research areas include monitoring AI reasoning processes, understanding social and emotional impacts, and evaluating AI's effects on economic systems through simulated real-world tasks.

The expanded partnership reflects Google DeepMind's commitment to responsible AI development alongside its broader safety strategy, which includes rigorous testing of models like Gemini 3 and collaboration with other external experts. This collaborative approach aims to build more robust safety frameworks that benefit not just the organizations involved, but the wider industry and everyone who interacts with AI systems.