The U.S. and China are expected to hold AI safety talks later this month. With AI capabilities improving fast, lots of people are calling for a grand bargain to prevent catastrophe. I'm also deeply concerned about the risks. But I think there's a more pragmatic way forward. In @ForeignPolicy, I make the case for sequencing smaller deals to enable progress. Each small agreement helps build the understanding and trust needed for more ambitious deals. This approach gives both sides options as the technology evolves. I think there's low-hanging fruit that can build on this approach: 1. Prevent extreme risks. Discuss best practices and begin carefully scoped technical exchange on safety evaluations and safeguards. 2. Manage crises. Create a voluntary information-sharing channel on critical incidents and establish a non-phone-based hotline. 3. Invest in verification tools. Verification is a nascent technology with lots of uncertainty. Nonetheless, both sides could begin to explore it by committing domestic R&D funding to expand the option space for future negotiation.

The new Burnham administration in the UK recognizes one of the core truths of politics: all politics are local. And UK AI policy will benefit from localization, including in areas like manufacturing capabilities. But there is another, less frequently asserted truth: that all local politics are structurally shaped by the international. Thinking geopolitically, and helping the UK navigate a fraught international environment, will be necessary if the country wants to be less susceptible to global shocks that would impede growth. In the global AI arena, Burnham inherits a remarkable hand. He has the world's most capable government: one that American companies rely on to evaluate their models before they are released publicly, one that the Chinese respect diplomatically due to the success of the Bletchley Summit, one that the world trusts to articulate the truth of the state of the science in the International AI Safety Report. There are excellent start-ups being built here that attract global talent; the UK's universities are producing world-class research in AI development, governance, and safety. It's been achieved through cross-party support and strong cross-party leadership. It is truly remarkable. All of this has allowed the UK to play a role bigger than its physical capital alone would predict. And that role is so very fragile. Now is the time for the UK to shore up these international advantages. If it doesn't, it risks imperiling both the domestic growth agenda as well as its global status. Luke Cavanaugh and I offer some initial thoughts on the pathway forward for @CarnegieEndow

As the two dominant powers in transformative AI, Washington and Beijing will determine whether it creates widely shared benefits or generates dangerous new risks. An AI model from either country could potentially be used to engineer a pathogen, launch cyberattacks, or create and disseminate realistic deepfakes — from anywhere in the world. In @ForeignAffairs, @cqknight_ and I argue that a prudent U.S. risk mitigation strategy does not mean slowing down innovation. Instead, it means working with Beijing to come to an understanding of safety research priorities, to coordinate testing for vulnerabilities and implementing safeguards, and to jointly establish best practices to contain truly global risks. Working together is necessary, and with the right approach, it is feasible. By focusing on how to look for risks rather than the specifics of what they find, Washington and Beijing can compete fiercely on AI while still mitigating the most extreme dangers it presents to the world.