Stanford OnlineWebinar: Regulating at the Speed of AI with Francis Fukuyama and Nate Persily
Stanford Online
Course focus and audience 0:10
Francis Fukuyama opens by describing a new in-person course, AI, power and the new rules of global strategy, planned for the week of March 22, 2027. He says it is not a skills class or a guide to using AI in business. It is for people interested in governance, geopolitics, domestic politics, and the international shape of the world AI is making. Nate Persily, a law professor and AI policy expert, joins him.
Regulation and enforcement 2:00
Persily says AI governance is already here, from tort law after a self-driving car crash to rules for banks and drug companies. He argues that regulating the technology itself is now a serious question, but the field is fast moving, international, and tied to China. He doubts frontier labs can safely police themselves after recent incidents, and he says any U.S. system needs real people, technical talent, and some kind of specialized regulator, while national security, energy, child safety, and other areas still need broader oversight.
China and regulation 17:01
The race frame is set aside. The point is not to win a finish line, but to manage what happens if recursive self-improvement makes models far more capable. That leaves room to keep investing in research while also making sure agents do not escape sandboxes. Open models from China also create a special problem, because they can be adapted worldwide and are harder to control with rules aimed only at closed systems.
Politics and work 23:02
The politics are already shifting. Data centers have become the concrete target for wider fears about AI, from kids and jobs to energy and water. A wider tech backlash also sits behind it. The economic effects may still be limited in the short term, but if large-scale job loss comes, the deeper issue will be redistribution, dignity, and whether millions of people can be satisfied with hobbies instead of paid work.
Different roles for countries 33:30
Some countries will not build their own foundation models. The better path is to think in layers. Brazil has abundant energy and a grid that is over 95 percent renewable, so it can host data centers. Japan and Europe can still build strong application businesses if they make the technology easier to spread through the economy.
Open models and regulation 35:02
The US, and maybe China and South Korea, will lead on chip design, while Taiwan remains central because TSMC can make semiconductors at a level others cannot. A key issue is trust. When the US cut off one model that others were using, it shook confidence. Brazil responded by building both American and Chinese supercomputing systems and hedging its bets.
Near-term AI harms 50:00
Even if the United States found the perfect regulation, it might not stop the worst cases. The sharper worry here is not a doomsday ending but a loss of control, with open-source agentic superintelligence posing a hard problem. More likely are smaller incidents, such as damage to the electrical grid or stock markets, alongside major benefits like better medicine and cancer cures.
AI-generated summary. It can be wrong or incomplete - check anything that matters against the original.
