How summits in Seoul, France and beyond can galvanize international cooperation on frontier AI safety
Last 12 months, the UK Government hosted the primary major global Summit on frontier AI safety at Bletchley Park. It focused the world’s attention on rapid progress on the frontier of AI development and delivered concrete international motion to reply to potential future risks, including the Bletchley Declaration; latest AI Safety Institutes; and the International Scientific Report on Advanced AI Safety.
Six months on from Bletchley, the international community has a chance to construct on that momentum and galvanize further global cooperation at this week’s AI Seoul Summit. We share below some thoughts on how the summit – and future ones – can drive progress towards a standard, global approach to frontier AI safety.
AI capabilities have continued to advance at a rapid pace
Since Bletchley, there was strong innovation and progress across the whole field, including from Google DeepMind. AI continues to drive breakthroughs in critical scientific domains, with our latest AlphaFold 3 model predicting the structure and interactions of all life’s molecules with unprecedented accuracy. This work will help transform our understanding of the biological world and speed up drug discovery. At the identical time, our Gemini family of models have already made products utilized by billions of individuals around the globe more useful and accessible. We have also been working to enhance how our models perceive, reason and interact and recently shared our progress in constructing the long run of AI assistants with Project Astra.
This progress on AI capabilities guarantees to enhance many individuals’s lives, but in addition raises novel questions that have to be tackled collaboratively in quite a lot of key safety domains. Google DeepMind is working to discover and address these challenges through pioneering safety research. Previously few months alone, we’ve shared our evolving approach to developing a holistic set of safety and responsibility evaluations for our advanced models, including early research evaluating critical capabilities akin to deception, cyber-security, self-proliferation, and self-reasoning. We also released an in-depth exploration into aligning future advanced AI assistants with human values and interests. Beyond LLMs, we recently shared our approach to biosecurity for AlphaFold 3.
This work is driven by our conviction that we want to innovate on safety and governance as fast as we innovate on capabilities – and that each things should be done in tandem, repeatedly informing and strengthening one another.
Constructing international consensus on frontier AI risks
Maximizing the advantages from advanced AI systems requires constructing international consensus on critical frontier issues of safety, including anticipating and preparing for brand new risks beyond those posed by present day models. Nonetheless, given the high degree of uncertainty about these potential future risks, there is obvious demand from policymakers for an independent, scientifically-grounded view.
That’s why the launch of the brand new interim International Scientific Report on the Safety of Advanced AI is a vital component of the AI Seoul Summit – and we stay up for submitting evidence from our research later this 12 months. Over time, such a effort could develop into a central input to the summit process and, if successful, we imagine it ought to be given a more everlasting status, loosely modeled on the function of the Intergovernmental Panel on Climate Change. This might be a significant contribution to the evidence base that policymakers around the globe must inform international motion.
We imagine these AI summits can provide a daily forum dedicated to constructing international consensus and a standard, coordinated approach to governance. Keeping a singular deal with frontier safety may also ensure these convenings are complementary and never duplicative of other international governance efforts.
Establishing best practices in evaluations and a coherent governance framework
Evaluations are a critical component needed to tell AI governance decisions. They permit us to measure the capabilities, behavior and impact of an AI system, and are a vital input for risk assessments and designing appropriate mitigations. Nonetheless, the science of frontier AI safety evaluations continues to be early in its development.
This is the reason the Frontier Model Forum (FMF), which Google launched with other leading AI labs, is engaging with AI Safety Institutes within the US and UK and other stakeholders on best practices for evaluating frontier models. The AI summits could help scale this work internationally and help avoid a patchwork of national testing and governance regimes which are duplicative or in conflict with each other. It’s critical that we avoid fragmentation that would inadvertently harm safety or innovation.
The US and UK AI Safety Institutes have already agreed to construct a standard approach to safety testing, a vital first step toward greater coordination. We expect there may be a chance over time to construct on this towards a standard, global approach. An initial priority from the Seoul Summit may very well be to agree a roadmap for a big selection of actors to collaborate on developing and standardizing frontier AI evaluation benchmarks and approaches.
It’ll even be vital to develop shared frameworks for risk management. To contribute to those discussions, we recently introduced the primary version of our Frontier Safety Framework, a set of protocols for proactively identifying future AI capabilities that would cause severe harm and setting up mechanisms to detect and mitigate them. We expect the Framework to evolve significantly as we learn from its implementation, deepen our understanding of AI risks and evaluations, and collaborate with industry, academia and government. Over time, we hope that sharing our approaches will facilitate work with others to agree on standards and best practices for evaluating the protection of future generations of AI models.
Towards a world approach for frontier AI safety
Lots of the potential risks that would arise from progress on the frontier of AI are global in nature. As we head into the AI Seoul Summit, and sit up for future summits in France and beyond, we’re excited for the chance to advance global cooperation on frontier AI safety. It’s our hope that these summits will provide a dedicated forum for progress towards a standard, global approach. Getting this right is a critical step towards unlocking the tremendous advantages of AI for society.
