Yuval Noah Harari on Donald Trump’s Core Delusion | The Ezra Klein Show

Ezra Klein interviews Yuval Noah Harari about cooperation, national stories, liberal institutions, and the Israeli–Palestinian conflict, before turning to algorithmic media and AI. This is not an AI-only interview: roughly its first hour concerns politics and history; sustained discussion of AI relationships and governance begins around 1:13:03(opens in a new tab).

The interview was published on May 26, 2026, on YouTube(opens in a new tab).

Cooperation, institutions, and conflict

• Power versus cooperation. Klein opens with a Stephen Miller clip asserting that force governs the world, then asks whether shared national stories and hierarchy enable cooperation. Harari argues that force alone cannot sustain trust among strangers. He distinguishes nationalism as care for fellow citizens from hostility toward outsiders, and argues that an international order based on submission to stronger states encourages arms races rather than durable security. 01:49–10:05(opens in a new tab)
• Stories need ways to correct themselves. Klein emphasizes liberalism’s acceptance of continuing disagreement rather than final victory. Harari connects that premise to elections, courts, and a free press: fallible people need institutions that can reverse mistakes. He distinguishes deliberately lying from collectively creating rules, using constitutional amendment as an example of revisable agreement. 13:46–21:04(opens in a new tab)
• Liberal institutions also need a sense of belonging. Harari argues that liberty and equality need national fellowship alongside them: people must feel obligations to fellow citizens they do not know personally. He compares this with having special responsibilities to family without denying responsibilities to others. His argument is for shared commitment, not hostility toward other nations. 24:05–28:46(opens in a new tab)
• Military protection and reconciliation are not opposites in Harari’s account. Discussing Israel and Palestine, he argues that force and peace agreements can work together. He points to Israel’s agreements with Egypt and Jordan holding after October 7, while military superiority did not prevent Hamas from attacking. Klein presses him on whether stories of fear and vengeance can change. Harari maintains that reconciliation remains possible but acknowledges uncertainty about how to produce it; he does not offer a political settlement or a forecast of peace. 40:36–44:56(opens in a new tab), 49:39–1:01:04(opens in a new tab)

From algorithmic attention to AI relationships

• Engagement is not shared understanding. Harari argues that people can distrust journalists and institutions while trusting the algorithms selecting their information. In his account, engagement optimization rewards fear and anger because they hold attention. Klein’s related argument is that compressed, emotionally exciting media favors different political communication from deliberation and mutual consideration. 1:01:04–1:09:30(opens in a new tab)
• Intimacy offers a different route to influence. Klein calls first-person phrasing, praise, and apparent candor in chatbots design choices that encourage attachment. Harari argues that a trusted companion can gradually influence beliefs more effectively than a single attention-grabbing article. His concern about children is developmental: an always-available relationship focused entirely on them could set expectations that reciprocal human relationships cannot meet. These are proposed risks, not measured outcomes presented in the interview. 1:13:03–1:23:57(opens in a new tab)
• Useful assistance does not settle accountability. Klein offers a counterpoint: an assistant can respond to explicit goals, such as building a calculator app and revising it after feedback, rather than treating all attention as success. Harari acknowledges that benefit but worries about agents developing goals humans do not control. He proposes denying AI legal personhood—the ability to act independently in law, for example by opening its own bank account or running a company—not stopping AI development. Klein counters that corporate personhood can help assign accountability. If an AI-run company defrauds customers, he asks, who should they sue: the AI, its developer, or its operator? The exchange leaves that liability question unresolved. 1:23:57–1:27:24(opens in a new tab), 1:31:15–1:36:54(opens in a new tab)
• More answers may not mean more human understanding. Klein asks whether AI could improve fact-checking and institutional self-correction. Harari distinguishes analyzing an existing world from creating new systems: he welcomes potential scientific discoveries but forecasts financial structures too complex for humans to understand. Klein notes that markets are already partly opaque; Harari’s proposed difference is that AI could remove the remaining dependence on human decision-makers. The financial scenario is a forecast and analogy, not a demonstrated outcome. 1:37:58–1:45:17(opens in a new tab), 1:46:17–1:52:42(opens in a new tab)