AI Power Shift | How the World Is Really Regulating AI

A frontier-lab researcher just quit publicly and said there's more than a 10% chance AI kills humanity within a decade, and his former colleagues said, "Yeah, he's right, we talk about it all the time." But that's not even the scariest part of this conversation. In this episode of the ProductMind Podcast, Oji Udezue and Ezinne Udezue (co-authors of Building Rocketships) and Ted Yang (author of Ageless: Peak Performance Using Artificial Intelligence) map how the world is actually approaching AI regulation, drawn from Ted's recent travels through Europe, Central and South America, and Ezinne's consulting work with clients across the EU. The picture that emerges is a global power shift most people building products aren't tracking: while the US outsources the hard questions to courts, corporate NDAs, and utility bills, Europe, China, and the Global South are each charting very different paths. The through-line the team keeps returning to: the real question isn't "data centers bad" or "privacy bad." We need data centers. We need to protect privacy. The question is what the good version of these things looks like, and right now, the only brakes on the whole system are the frontier labs' own conscience, weighed against their fiduciary duty to increase shareholder value. That's the argument for external governance, whatever you want to call it. What we cover in this episode: Why European startups want regulation (privacy as a fundamental right, and protection against frontier labs) • The XUS investment strategy, countries funding home-grown AI to escape US dependency • How Trump-era tariffs moved up the world's timeline to decouple from America • The EU AI Act; watermarking, transparency duties, and the pullback after industry pushback • China's "algorithmic filing"; full documentation of how models are trained, and what that proves is possible • Why the average person in China is pro-AI while Americans are skeptical • The UK's principles-based approach and why Canada's AI act died in parliament • Harvard's "data colonialism" framing; extracting data from the Global South and selling it back as inference • Data-labeling arbitrage, the Kenya health-data deal, and techno-colonialism • Why the US response is "pure laziness and absolving themselves" • The token math: why it takes 300M–1B tokens per person to move a nation's productivity and why this isn't a bubble • Regulatory capture vs. real governance • The public frontier-lab resignation and recursive self-improving models • What product people should actually do now: model selection, data hygiene, being a good citizen as differentiation About ProductMind: ProductMind explores how AI is reshaping products, work, and society. We talk about the hard technology and the human side; subscribe for candid conversations with the people building and thinking hard about the future. 🔔 Subscribe for new episodes on AI, regulation, product, and the future of work. 💬 Drop a comment: Should AI be treated as a common good? Where does your country sit on this map and who should hold the brakes? #AIRegulation #AIPowerShift #ArtificialIntelligence #EUAIAct #DataColonialism #ProductMind #aigovernance #chinaai #TedYang #BuildingRocketships #aiandsociety #futureofwork #datacenters #RegulatoryCapture #globalsouth