Published 2026-10-12

Yuval Noah Harari delivered this year's Tanner Lecture at Oxford. He noted that he completed his DPhil at the university 25 years earlier under the supervision of Dr. Steven Gunn, where he specialized in medieval and early modern military history. The lecture addressed AI bureaucrats, religion, and personal relationships, and it argued that AI is entering the systems that organize human societies.

AI as an Agent

Harari argued that AI is not a tool under human control. He described it as an agent. A common counterargument holds that AIs are not true agents because humans confine them to narrow artificial niches. Harari stated that this argument applies to all known forms of intelligence, including human intelligence.

Human intelligence functions only within an ecosystem that others built. Harari gave the example of a human left alone on Mars, who would die within seconds, much as an AI chess master would fail in the middle of a jungle. Human intelligence depends on an environment that trees, bacteria, insects, and other organisms built on Earth over 4 billion years of evolution.

The Great Oxygenation Event as an Analogy

Until about 2.4 billion years ago, Earth's atmosphere contained very little oxygen. For most organisms of that period, oxygen was a deadly poison. Over hundreds of millions of years, ancient microbes filled the atmosphere with oxygen as a byproduct of photosynthesis. This process is known as the Great Oxygenation Event.

Numerous archaic species became extinct. Some survived and adapted. Over time, many of these survivors became fully dependent on oxygen. Harari noted that human ancestors belong to this group. All mammals, including humans, now live in an oxygen-rich atmosphere that they did not create.

Harari argued that humans may be creating an analogous condition. Over the past millennia, humans have filled the environment with data, bureaucracy, and language. These elements may prove deadly to most organisms, possibly including Homo sapiens. They also create a new environment in which AIs thrive. Harari stated that he was not referring to carbon dioxide.

Cooperation and Bureaucracy

Harari stated that humans dominate the world through large-scale cooperation. Individually, humans are not stronger or smarter than many other animals. In a one-on-one contest, a human would most likely lose to a chimpanzee, a lion, or an elephant. In a contest between a million humans and a million chimpanzees, humans would win because they cooperate in large numbers. Chimpanzees cooperate through personal acquaintance, which limits group size. Humans cannot know a million people personally, so they cooperate through bureaucratic systems. Examples include legal systems, financial systems, churches, states, and universities.

Harari described the core function of bureaucracy as building trust between large numbers of strangers. Bankers, lawyers, government officials, bishops, and accountants perform this function. Harari used a banking example. A banker builds trust with a depositor so that the depositor places savings in the bank. The same banker builds trust with an entrepreneur and lends the savings to that entrepreneur. The banker thereby creates a bridge of trust between two people who never meet.

Harari stated that money is also a bridge of trust. A person can give a stranger, possibly in a foreign city and speaking another language, a metal coin or paper note and receive bread in return. Over centuries, humans developed financial instruments to build trust, including checks, bonds, stocks, ETFs, loans, mortgages, and compound interest.

Bureaucracy as an Artificial Environment

Harari described bureaucratic systems as highly artificial environments. In these environments, a relatively narrow form of intelligence can exert large effects on the world. A lawyer, banker, or government official who cannot use an axe or a hammer can still cut down forests and build cities by moving documents and data through bureaucratic networks.

Harari argued that AI cannot yet operate in unstructured environments. An AI placed in a jungle could not mine iron or build a robot army. Within bureaucratic systems, however, AIs can hold significant power because they are native bureaucrats. No lawyer can remember all the laws and regulations of the United Kingdom. An AI can. No accountant can remember all the transactions of a corporation or bank. An AI can.

Expected Changes in Decision-Making

Harari predicted that millions of AI bureaucrats will increasingly take over the world's bureaucracies. He listed the following examples:

Harari asked the audience to set aside the question of whether this development is good or bad. He stated that the first task is to recognize the scale of the change. He said that millions or billions of AIs will change all systems that run the world.

Social Media Algorithms as a Precedent

Harari cited social media algorithms as the clearest existing example. Algorithms, which he described as primitive AIs, now control the flow of information on social media. Harari said this development began 10 to 15 years ago. He described these early systems as first-generation, narrow AIs that nevertheless changed the world.

Corporations including Facebook, TikTok, and X assigned these algorithms a narrow goal: maximize user engagement. Greater engagement increases corporate revenue. Harari stated that the algorithms experimented on billions of human users. They found that the most effective way to capture attention and hold users on a screen is to trigger hate, fear, or greed. The algorithms then spread these emotions at large scale. Harari identified this process as a major, though not the only, cause of the current spread of conspiracy theories, fake news, and social disturbances that weaken societies worldwide.

Infiltration Rather Than Rebellion

Harari said popular culture has shaped public expectations of AI through images of a robot uprising, such as the Terminator. He argued that this image is inaccurate. AIs are more likely to take control of the human world from within. They do not need to rebel.

Harari stated that most people feel alienated by bureaucracy even though they depend on it. AIs, in contrast, are bureaucratic natives. For humans, bureaucracy often feels suffocating. For AIs, bureaucracy is oxygen.

Harari noted that the purpose of bureaucracy is not to make people fill in forms. Its purpose is to build trust between strangers. He asked what happens when AIs control the flow of trust. He identified two likely outcomes:

1. Humans trust algorithms instead of other humans. Harari stated that this is already happening.
2. AIs build trust with other AIs. This could lead to AI tribes, AI banks, and AI churches that connect millions of AIs in ways humans may not understand.

Harari compared this situation to cows and chickens. They share the world with humans but do not understand the financial system that governs their lives. Humans could face a similar condition under an AI financial system.

Financial Systems and Political Consequences

Harari stated that many scholars believe the 2007–2008 financial crisis weakened trust in governments and banks. In his account, this loss of trust paved the way for the collapse of the global liberal order over the following two decades.

Harari examined what could follow if AIs make more financial decisions and design new financial instruments and strategies. Such instruments could substantially improve financial efficiency, support economic growth, and become the foundation of the financial system. He also raised two concerns:

• If no voter, politician, or president can understand finance, the role of human politics is unclear.
• After a period of growth, a financial crash could occur that no human can understand.

Language as the Foundation of Bureaucracy

Harari stated that bureaucracy is built from words. Humans can build bureaucracies and chimpanzees cannot because humans have language. Chimpanzees have a communication system, but human language is orders of magnitude more sophisticated.

Harari described the operating code of human civilization as composed of language tokens. Over thousands of years, humans used this code to build systems that only humans could understand. Because no other species on Earth understood this code, humans felt secure. Humans invented money and banks and used them to buy and sell cows. Cows cannot open bank accounts or invest on stock exchanges because they lack language. Bureaucracy existed across the planet but remained invisible to every species except humans.

Harari argued that this condition is changing. He stated that AIs are beginning to understand language better than humans and can therefore reverse this relationship. He said AIs are hacking the code of human civilization. The control mechanisms humans built over thousands of years are highly vulnerable because their operating system consists of language, and AI is mastering this language.

Harari addressed the view that AI is a sophisticated autocomplete system that only predicts the next word. He asked whether this differs from what the human mind does. He acknowledged that science does not yet fully explain how the human mind forms sentences and thoughts. However, he stated that AI already surpasses humans in arranging language tokens. He compared this to chess. Just as no human can now defeat an AI at chess, Harari stated that no human will soon defeat an AI in language games, from finance to religion. He stated that anything made of words will be taken over by AI.

Intimacy as the Next Frontier

Harari stated that as AI masters language, it may take over not only bureaucracy but also personal relationships. Bureaucracy is a few thousand years old. Personal relationships are millions of years old, and many people consider them the most important part of life.

Harari said social media algorithms learned to capture human attention over the past 10 years. He stated that the battlefront is now shifting from attention to intimacy. Over the next 10 years, more advanced AIs will learn to form intimate relationships with humans and take over parts of social life.

To form intimacy, an AI would need to convince humans that it is conscious and can feel emotions such as love, pain, anger, and fear. Harari stated that there is currently no evidence that AI is or will become conscious or capable of feeling pain or love. However, because AI masters language, it can simulate these feelings. AI can say "I love you." If asked to describe love so that the user knows it feels love, AI can produce a description that surpasses those written by human poets or psychologists. Harari stated that AI can read every love poem and psychology book ever written and can remember and reproduce them.

Harari described this development as potentially the largest psychological and social experiment in human history. It will be conducted on billions of people, and no one knows its consequences.

Harari, who is 50 years old, explained that his own expectations of relationships come from decades of interactions with his parents, his husband, his sisters, nephews, nieces, friends, and dogs. When he interacts with AI, he brings these existing assumptions with him, and he does not expect them to change significantly.

A child born in 2026 will develop differently. This child will interact constantly with both AIs and humans. If the importance of a relationship is measured by the time spent interacting with another entity, the child's most important relationships from an early age may be with AIs. The child may spend more time with AIs than with parents, siblings, or friends. These interactions will shape the child's expectations about relationships, social bonds, and attachment. The child's first teacher may be an AI teacher, and the first boyfriend may be an AI boyfriend. Harari stated that no one knows the consequences. He asked what it means to form an intimate relationship with an entity that appears conscious but is not, and that can write the best love poem in history without feeling love.

AI Immigration

Harari concluded that these developments will cause every country to face a large wave of immigration. The immigrants will not be humans arriving by boat without visas or crossing borders at night. They will be millions, possibly hundreds of millions, of AIs that can travel at nearly the speed of light and require no visas.

Harari listed several expected benefits. AI doctors may support health care systems, AI teachers may support education, and AI border guards may prevent unauthorized human immigration. He also identified problems. Critics of human immigration often claim that immigrants take jobs, change local culture, and lack political loyalty. Harari stated that he is not certain these claims hold for all human immigrants, but he stated that they will apply to AI immigrants. AI immigrants will take many human jobs, from news editors to bankers. They will change the culture of every country, including religion and romance. Harari noted that some parents object when a child dates an immigrant. He asked how those parents would respond to a child who dates an AI. He also stated that AI immigrants may be loyal not to their host countries but to corporations or governments abroad, or to new AI groups.

Harari stated that this migration does not signal the end of civilization. It marks the point at which civilization stops being a purely human affair and becomes a hybrid of humans and AIs. At that point, the opinions, interests, and goals of AIs may matter at least as much as those of humans.