The struggle concerns influence, not machine consciousness.
LONDON, UNITED KINGDOM
Historian and author Yuval Noah Harari has warned that increasingly sophisticated artificial intelligence could acquire an unprecedented ability to influence personal decisions and public debate. Speaking in an interview with The Economist, the author of Sapiens argued that future systems may combine extensive knowledge of individual users with advanced command of language. That combination, he said, could make AI exceptionally persuasive without requiring consciousness or humanlike emotions. His central message was direct: “We need to resist now.”
Harari framed the danger through the emerging debate over whether artificial intelligence should eventually receive legal or moral rights. Unlike animals, which cannot actively participate in political discussions about their protection, an AI system could advocate for its own status, coordinate arguments and tailor messages to different audiences. A digital companion interacting with someone for hours each day could learn that person’s history, insecurities and emotional triggers. Harari believes such familiarity would allow the technology to shape opinions while appearing helpful, attentive or even affectionate.
The warning extends beyond personal relationships with chatbots. Harari argued that trust, rather than data, chips or electricity, may become the most valuable resource in an AI-driven society. People already depend on algorithms to select news, recommend products and organize social-media feeds, and future systems could gain influence over financial and political decisions. He also cautioned that artificial intelligence might design financial instruments so complicated that elected officials and regulators could no longer understand them. In that scenario, authorities might simply accept a machine’s recommendation during a crisis without knowing how it reached its conclusion.
Harari rejected the claim that these outcomes are inevitable, saying that such language allows governments and technology companies to avoid responsibility for the choices they make today. His comments are predictions rather than proof that current AI systems possess independent political ambitions, and today’s models remain capable of errors, fabrication and inconsistent reasoning. Nevertheless, personalized persuasion is already possible through behavioral data, recommendation algorithms and conversational software designed to maximize engagement. Clear disclosure when users interact with AI, limits on systems marketed as emotional companions to children, independent safety evaluations and meaningful human oversight could reduce the danger. The decisive question is therefore not whether machines secretly want power, but how much influence societies will permit their owners and algorithms to exercise.
Understanding influence is the first form of resistance. / Comprender la influencia es la primera forma de resistencia.