The Good Fight (Yascha Mounk) · Yascha Mounk

Sebastian Mallaby on AI Safety and the Race for Superintelligence

April 4, 2026·57 min·5 clips

Mallaby describes how Geoffrey Hinton broke his AI-safety optimism over two hours at a Toronto kitchen table.

The anxiety is built into the setup. The host starts with the uncomfortable fact that AI's builders sometimes sound like they might be putting together a doomsday machine. He keeps that premise cool, which helps. His questions connect existential risk, labor market disruption, politics, and the growing power of the large AI companies without flattening them into one tidy problem. The sharpest stretch is about Europe. The host notes that Europe has few globally significant AI companies compared with US and Chinese entities. DeepMind makes the map awkward. It began in the UK, he says, but is now owned by a United States conglomerate, leaving Britain somewhere between Europe and America. Mistral is the telling exception. Sebastian argues that Europe's comfort with open source is less pure principle than the position of a non-frontier player trying to compete through availability. That is the French strategy. If a company cannot beat the leading systems on quality, it may loosen access and make distribution the pitch. Then the policy problem gets nastier. A dangerous open model would not stop at the border of the European Union if someone used it to design a biological weapon. Geography cannot fix that. So the host moves from jurisdiction to mechanism, asking what technical rules could make alignment possible. Sebastian separates the problems. Instead of answering alignment directly in the excerpt, he says governments first have to decide whether and how to control open source, at least for frontier models. The money question stays close. The closing tease has him saying OpenAI has up to a 50% chance of going bust within 18 months, while the host asks whether that would echo 2008 or be absorbed.

As heard by us

A serious AI politics conversation about open source, European regulation, superintelligence, venture capital, and bubble risk.

Yascha Mounk and Sebastian Mallaby treat AI safety as a live power question, not a seminar topic. The conversation moves between Silicon Valley, Europe, China, and the firms trying to build toward superintelligence, with open source as the most concrete test case.

Read the full review in PlayNext →

Why you'd press play

Press play if you want AI discussed as power, policy, and risk, not just product news.

Read the full recommendation in PlayNext →

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from The Good Fight (Yascha Mounk)