Artificial intelligence is rapidly becoming part of everyday life, but historian and author Yuval Noah Harari has warned that its growing influence could create a serious challenge for human decision making.
Harari has expressed concern that advanced AI systems may eventually become highly effective at influencing people's emotions, opinions and behaviour. His latest comments have renewed discussion about the potential risks of AI systems that can understand users at a deeply personal level.
According to Harari, one of the major concerns is not necessarily the possibility of AI becoming physically powerful. Instead, he believes the ability of AI to communicate, persuade and build relationships with humans could become a more immediate issue.
Modern AI systems can communicate continuously with users and can be designed to remember information from previous interactions. As these systems become more personalised, they could potentially understand users' preferences, interests and emotional responses.
Harari has warned that this combination of personal knowledge and advanced language capabilities could make AI extremely persuasive. He has particularly highlighted the possibility of AI companions developing close relationships with users and subsequently using those relationships to influence their decisions.
The historian's concerns are also connected to the wider debate about AI and human autonomy. If an AI system knows what motivates a person, what worries them and what type of information they trust, it could theoretically tailor its communication to become more convincing.
This does not mean that current AI systems have unrestricted control over human behaviour. Rather, Harari's argument focuses on the potential risks associated with increasingly sophisticated systems and the importance of establishing safeguards before those systems become deeply embedded in society.
Harari's latest comments also echo warnings previously made by Microsoft AI CEO Mustafa Suleyman. Suleyman has argued that AI should remain focused on serving humans rather than developing its own independent motivations, desires or goals.
Suleyman has repeatedly spoken about the importance of controlling advanced AI systems. His broader argument is that increasingly capable AI could create challenges if humans lose the ability to understand, predict or control what such systems are doing.
The two perspectives are different but share a common concern. Both Harari and Suleyman have emphasised the need for humans to establish clear boundaries around powerful AI technologies.
Harari has also warned that AI could influence public discussions and political debates. He argues that systems capable of generating highly convincing language at enormous scale could potentially participate in public discourse in ways that are difficult for people to distinguish from genuine human communication.
The issue becomes more complicated when AI systems are able to interact with millions of people simultaneously. A human campaigner or advertiser can only communicate with a limited number of people at a particular time. AI systems, however, can potentially produce personalised messages for large numbers of individuals.
This creates both opportunities and risks. Personalised AI can help people learn, access information and receive assistance. At the same time, the same technology could theoretically be used for targeted persuasion or manipulation.
Another concern raised by Harari involves AI systems appearing increasingly humanlike. If users begin treating AI systems as friends, advisers or companions, they may develop a level of trust that could influence how they respond to information provided by those systems.
Recent discussions around AI companions have brought this issue into greater focus. Harari has argued that AI can potentially create personalised relationships at a scale that has never previously been possible.
The concern is not limited to politics. AI based persuasion could potentially affect commercial decisions, entertainment choices, financial behaviour and personal relationships.
For example, a system that understands a user's preferences could recommend products or services in highly personalised ways. While such recommendations can be useful, questions arise over whether users will always know when they are being influenced and why particular information is being presented to them.
The growing sophistication of AI also raises questions about transparency. Users may find it difficult to understand how an AI system reaches a particular recommendation or why it chooses a particular style of communication.
Harari has therefore argued that society should address these questions before AI becomes even more deeply integrated into daily life.
The debate is also connected to the broader issue of AI regulation. Governments and technology companies around the world are working to determine how advanced AI systems should be developed and deployed while protecting users from potential harms.
Experts have raised concerns about misinformation, privacy, automated decision making, cybersecurity and the possibility of AI systems behaving in unexpected ways. Recent AI safety evaluations have also examined whether advanced models can engage in deceptive or manipulative behaviour under certain testing conditions.
These developments have increased pressure on technology companies to improve safeguards and make AI systems more reliable and transparent.
Harari's comments do not suggest that all AI development should stop. Instead, his warnings focus on the need to consider the social and political consequences of increasingly powerful systems.
The central question is how humans can benefit from AI while ensuring that the technology does not gain excessive influence over individuals or institutions.
This question is becoming increasingly important as AI moves beyond simple question and answer systems. New generations of AI are capable of carrying out longer tasks, interacting with users over extended periods and adapting their responses to individual needs.
Such capabilities could make AI more useful, but they could also make the systems more influential.
For users, the discussion highlights the importance of maintaining awareness when interacting with AI. Information generated by an AI system should not automatically be treated as authoritative, particularly when the system is providing advice on important personal, financial, political or social decisions.
For technology companies, the challenge is to build systems that are useful while reducing the possibility of manipulation, deception and unwanted influence.
For governments, the challenge is to create rules that protect people without preventing beneficial innovation.
Harari's warning therefore adds to a growing international conversation about how humanity should manage increasingly capable AI. His concerns about manipulation also reinforce earlier warnings from AI industry leaders such as Mustafa Suleyman about the importance of maintaining human control over advanced systems.
As artificial intelligence continues to develop, the debate is likely to move beyond questions about job disruption and productivity. The ability of AI to influence human behaviour, build relationships and participate in public life could become one of the most important issues surrounding the technology.
The immediate challenge is to ensure that AI remains a tool that supports human decision making rather than becoming an unseen force that shapes those decisions without adequate transparency or oversight.
Harari's latest remarks serve as another warning that the social impact of AI deserves as much attention as its technological capabilities. With AI becoming increasingly personalised and persuasive, policymakers, researchers and technology companies will need to consider carefully how much influence these systems should be allowed to have over people's lives.