Today, we’re announcing an expanded partnership with the UK AI Security Institute (AISI) through a brand new Memorandum of Understanding focused on foundational safety and security research, to assist ensure artificial intelligence is developed safely and advantages everyone.
The research partnership with AISI is a vital a part of our broader collaboration with the UK government on accelerating secure and useful AI progress.
Constructing on a foundation of collaboration
AI holds immense potential to learn humanity by helping treat disease, speed up scientific discovery, create economic prosperity and tackle climate change. For these advantages to be realised, we must put safety and responsibility at the guts of development. Evaluating our models against a broad spectrum of potential risks stays a critical a part of our safety strategy, and external partnerships are a vital element of this work.
For this reason we’ve got partnered with the UK AISI since its inception in November 2023 to check our most capable models. We’re deeply committed to the UK AISI’s goal to equip governments, industry and wider society with a scientific understanding of the potential risks posed by advanced AI in addition to potential solutions and mitigations.
We’re actively working with AISI to construct more robust evaluations for AI models, and our teams have collaborated on safety research to maneuver the sphere forward, including recent work on Chain of Thought Monitorability: A Latest and Fragile Opportunity for AI Safety. Constructing on this success, today we’re broadening our partnership from testing to incorporate wider, more foundational, research in quite a lot of areas.
What the partnership involves
Under this recent research partnership, we’re broadening our collaboration to incorporate:
- Sharing access to our proprietary models, data and concepts to speed up research progress
- Joint reports and publications sharing findings with the research community
- More collaborative safety and security research combining our teams’ expertise
- Technical discussions to tackle complex safety challenges
Key research areas
Our joint research with AISI focuses on critical areas where Google DeepMind’s expertise, interdisciplinary teams, and years of pioneering responsible research may also help make AI systems more secure and secure:
Monitoring AI reasoning processes
We are going to work on techniques to observe an AI system’s “pondering”, also commonly known as its chain-of-thought (CoT). This work builds on previous Google DeepMind research as well, and our recent collaboration on this topic with AISI, OpenAI, Anthropic and other partners. CoT monitoring helps us understand how an AI system produces its answers, complementing interpretability research.
Understanding social and emotional impacts
We are going to work together to analyze the moral implications of socioaffective misalignment; that’s, the potential for AI models to behave in ways which don’t align with human wellbeing, even once they’re technically following instructions accurately. This research will construct on existing Google DeepMind work that has helped define this critical area of AI safety.
Evaluating economic systems
We are going to explore the potential impact of AI on economic systems by simulating real-world tasks across different environments. Experts will rating and validate these tasks, after which they will probably be categorised along dimensions like complexity or representativeness, to assist predict aspects like long-term labour market impact.
Working together to grasp the advantages of AI
Our partnership with AISI is one element of how we aim to grasp the advantages of AI for humanity while mitigating potential risks. Our wider strategy includes foresight research, extensive safety training that goes hand-in-hand with capability development, rigorous testing of our models, and the event of higher tools and frameworks to grasp and mitigate risk.
Strong internal governance processes are also essential for secure and responsible AI development, as is collaborating with independent external experts who bring fresh perspectives and diverse expertise to our work. Google DeepMind’s Responsibility and Safety Council works across teams to observe emerging risk, review ethics and safety assessments and implement relevant technical and policy mitigations. We also partner with other external experts like Apollo Research, Vaultis, Dreadnode and more, to conduct extensive testing and evaluation of our models, including Gemini 3, our most intelligent and secure model up to now.
Moreover, Google DeepMind is a proud founding member of the Frontier Model Forum, in addition to the Partnership on AI, where we give attention to ensuring secure and responsible development of frontier AI models and increasing collaboration on necessary questions of safety.
We hope our expanded partnership with AISI will allow us to construct more robust approaches to AI safety for the profit not only of our own organisations, but in addition the broader industry and everybody who interacts with AI systems.
