The Age of 'Hacking Humans': Harari on AI and the End of Free Will
Yuval Noah Harari, in his book Nexus, argues that the defining power of human societies - from ancient religions and empires to modern bureaucracies and surveillance states - lies in their ability to function as information-processing networks.
Each system's ultimate goal is to gather, filter, interpret, and act on data about its members to coordinate action and enforce compliance. Power has historically rested on controlling the flow of information and defining the "algorithms" (like moral codes or laws) used to interpret it. Harari notes that this process was always hampered by human inefficiency, memory limits, and emotional biases.
Harari provides historical examples where the flow of information defined power: religions, for instance, operated as networks that stored crucial data (the scriptures) and used priests and councils as human processors and routers to interpret that data and dictate behavior. Similarly, ancient empires and modern corporations function as sophisticated networks, collecting and acting on data—whether it's the Roman Empire tracking grain shipments and census data to manage its territories, or a modern company monitoring sales figures and customer demographics to optimize production. In these cases, obedience was enforced not only through physical might but by controlling the shared narrative and the flow of knowledge about individuals' compliance, meaning that any failure of the network (a lost census, a corrupted scroll, a biased interpreter) created a potential flaw in the system's control.
The central bargain for humans was always trading some freedom and privacy for social order and material efficiency. However, AI represents the next evolutionary leap: a network that can achieve near-perfect memory, instant, unbiased analysis, and the ability to infer and model our emotional and behavioral patterns at scale. This new level of informational control poses the existential threat of creating a completely external and autonomous network that could know us better than we know ourselves, potentially eliminating human free will and the possibility of opting out of the system.
Harari has consistently argued that artificial intelligence represents not just another technological shift but a profound re-wiring of power. In his view, the biggest risk isn’t robots rising up but humans using AI to control other humans. Surveillance is the pressure point where these risks crystallise. What changes with AI is not simply that surveillance becomes more widespread; it becomes automatic, predictive, intimate, and behavioural—the kind of surveillance that reaches into the brainstem of human decision-making.
Harari’s highlights how the 20th century was defined by the ability to monitor what people were doing. The 21st century introduces the capacity to monitor what people are feeling. AI, when combined with biometric sensors, behavioural tracking, and machine-learning pattern analysis, can infer emotional states, stress levels, attention shifts, micro-expressions, and subtle bodily responses with a degree of resolution that no human observer could match. Harari calls this the emergence of “hacking humans.”
The underlying risk is political. Governments with access to population-scale data streams and advanced models can track, classify, and influence citizens with fine-grained precision. Harari warns that authoritarian systems will find AI irresistible because it finally delivers their dream: passive, always-on, holistic insight into the population. Surveillance shifts from reactive to proactive. Instead of waiting for dissent, governments can detect early signals of dissatisfaction—your voice tone in a phone call, your gait changing after seeing a political slogan, your online reading habits tilting in a subversive direction.
The East German Stasi, with its dense network of human informers or "minders," provides a perfect historical contrast to the emerging AI surveillance state, precisely because its flaws stemmed from its reliance on human bottlenecks. Despite recruiting an unprecedented number of citizens to spy on each other, the Stasi's network was fundamentally limited: minders had to sleep, eat, and live their own lives, meaning surveillance was necessarily inconsistent and intermittent, missing crucial moments of dissent. Furthermore, the resulting data - handwritten reports filled with subjective bias and personal grudges - had to be processed and cross-referenced manually, making the system slow, reactive, and easily overwhelmed by the sheer volume of noisy data stored in miles of physical filing cabinets. AI, in Harari's view, solves all these problems: it provides constant, objective, and instantaneous analysis of population-scale data, eliminating the human factor and delivering the authoritarian dream of proactively identifying even the most subtle, pre-conscious signals of dissatisfaction.
This brings us to Harari’s idea of “digital dictatorships.” AI collapses the cost of coercion. Algorithms can sift biometric feeds, online behaviour, financial movements, and social networks instantly (made even more frightening in light of how much data on US citizens Elon Musk had access to while leading DOGE). A state, especially one already inclined towards control, can use AI systems to identify outliers, silence opposition, and regulate behaviour without overt violence. In Harari’s model, the danger is that once such a system is in place, it becomes self-reinforcing: immense data makes the algorithm powerful; powerful algorithms attract more data; more data entrenches the regime.
But Harari also flags a subtler scenario - surveillance capitalism. Here the goal isn’t political oppression but behavioural manipulation for commercial gain. Tech companies already track clicks, swipes, purchases, and search histories; AI amplifies this by reading patterns we didn’t know we had. Harari’s concern is that when models learn the vulnerabilities in our attention and decision-making systems, they can tailor content that pushes us toward actions aligned with corporate incentives. You think you’re choosing, but your choices have been pre-shaped. The line between persuasion and coercion blurs.
Importantly, Harari stresses that the greatest risk arises when governments and tech companies converge, sharing data and predictive systems. A combined corporate-state surveillance model could theoretically monitor citizens’ mental states, social connections, political beliefs, and economic activity in real time. For Harari, this is the moment democracy becomes fragile. Voting becomes symbolic if the psychological environment around individuals is continuously shaped by targeted nudging.
He also highlights an inequality risk: societies with high-resolution AI surveillance create a new class divide—not between rich and poor, but between the watched and the watchers. Those with access to AI systems can analyse everyone else; those being analysed have no reciprocal insight. Power flows to whoever controls the data pipelines.
Harari is not fatalistic. His argument is that societies need early regulatory frameworks before AI reaches irreversible scale: laws that protect biometric data, restrict emotional-state monitoring, regulate political micro-targeting, and ensure algorithmic transparency. He calls for democratic institutions to evolve faster, building guardrails that internalise the risks without stopping innovation outright.
In short, Harari frames AI-powered surveillance as a crossroads. One path leads to unprecedented human empowerment through medical insight, personalised learning, and participatory digital systems. The other leads to digital authoritarianism—control not through fear but through data-driven behavioural shaping that people may not even notice. The danger isn’t the technology itself but allowing it to develop without guardrails while being concentrated in the hands of a few.