My wife once bought me a pair of sneakers after seeing them in a boutique. Before the day ended, advertisements for the same product began appearing on my Facebook timeline. I had not searched for the sneakers online or discussed them in any online conversation. Yet there they were, as though someone had been watching.
A friend described a similar experience. He had been discussing a particular brand with someone in a car. Later, when he visited a website, Google AdSense began displaying advertisements for the product.
Such experiences raise an unsettling question: how much do the technologies surrounding us know about our interests, habits and intentions, and how much can they infer without our conscious participation?
There are conventional explanations for targeted advertising, including browsing behaviour, location data, tracking technologies and information collected from connected devices. An advertisement appearing after a conversation does not, by itself, prove that a phone was listening. Nor does an eerily accurate recommendation mean that artificial intelligence can read minds.
But technology no longer needs to know everything about us to make remarkably accurate predictions about what we might want, believe or do next.
That possibility brings to mind George Orwell’s Nineteen Eighty-Four, the 1949 novel in which a totalitarian state watches its citizens, controls information and punishes independent thought. Its enduring symbol is Big Brother, the authority whose power rests partly on the knowledge that people may be watched at any time.
Orwell’s imagined year was 1984. What might his warning look like in 2084, when artificial intelligence could be far more powerful than the systems we use today?
The danger may not be a machine independently taking control of society. It could be the concentration of technological power in the hands of governments, corporations and political actors able to collect information, interpret human behaviour and influence the choices people make. AI would be the instrument; the people directing it would determine how that power was used.
Nigeria has already encountered an early version of this problem through political campaigning.
Cambridge Analytica, the political consultancy notorious for its use of personal data, worked for Goodluck Jonathan’s 2015 re-election campaign against Muhammadu Buhari. Reporting on the campaign described efforts to obtain information about Buhari.
Buhari won the election. Cambridge Analytica did not determine the outcome. But the controversy exposed a serious concern. Personal information and targeted messages could be used to influence voters in ways that the wider electorate might never see.
Artificial intelligence could take this approach much further.
An AI-powered operation could generate thousands of campaign messages tailored to individual fears and concerns. One voter might receive material designed to amplify anxiety about insecurity; another might see messages intended to deepen distrust of a political opponent. Deepfake videos, cloned voices and automated accounts could spread falsehoods at speed, damaging reputations and eroding confidence in elections before the truth catches up.
The danger is not that technology will decide every election. It is that those who control it could gain powerful new ways to influence what people believe, often without their realising how precisely a message was designed for them.
But what happens when technology moves beyond studying our online behaviour to interpreting signals from inside the human brain?
The prospect sounds like science fiction, yet researchers have demonstrated that AI can reconstruct aspects of language from brain activity. The potential is both hopeful and unsettling.
Consider someone who has lost the ability to speak following a serious illness or injury. The thoughts are still there, but communicating them may be painfully difficult or impossible. A brain-computer interface that translates neural signals into words could allow that person to speak to family, express preferences and make decisions independently.
Now imagine information derived from brain activity being sought for a very different purpose. What if an employer, insurer or government wanted access to information a person had never chosen to disclose? The technology that might restore someone’s voice could also raise difficult questions about who gets access to the most intimate information about them.
These are not equivalent uses, and current technology cannot freely extract anyone’s thoughts. But the contrast shows why the direction of this research matters as much as its scientific achievements.
In a 2023 study published in Nature Neuroscience, researchers described a non-invasive semantic decoder that used functional magnetic resonance imaging (fMRI) and AI to reconstruct sequences of words capturing the meaning of stories participants heard or imagined. The system could also interpret aspects of silent videos. It did not produce a perfect transcript of private thoughts, and participants had to cooperate both to train and to use the decoder.
The experiment required specialised scanning equipment and extensive individual training. It was not a machine capable of secretly reading a stranger’s mind. Nevertheless, it demonstrated that AI could recover aspects of language and meaning from brain activity, including when participants imagined a story rather than listened to one.
As Saturnin Pugnet explores in Phase Lock, advances in this field invite us to consider a future in which the boundary between private thought and information accessible to technology may become harder to defend.
The link to political manipulation is clear. Campaigns already use personal data to predict which messages might change voters’ opinions. Brain-computer interfaces raise the possibility of accessing some information generated by neural activity. In both cases, the central concern is who controls the information and what they are permitted to do with it.
The benefits of AI and brain-computer interfaces should not be dismissed. These technologies could transform medical treatment and help people with disabilities communicate and gain greater independence. The challenge is to protect those benefits while preventing coercive data collection, misuse of sensitive information and decisions made without meaningful consent.
Governments must set enforceable limits on how personal and neural information can be collected and used. Political campaigns must be held accountable for deceptive uses of personal data and AI-generated material. People must retain meaningful control over whether sensitive information about them is gathered and shared.
Orwell’s warning remains relevant because control does not always announce itself. It can grow through systems people accept for convenience, institutions they trust and technologies whose consequences become clear only after they are embedded in daily life.
Twenty Eighty-Four is not a prediction that AI will take control of humanity by 2084. It is a warning about what can happen when those who wield technological power gain the ability to monitor, influence and manipulate others faster than society can establish safeguards against abuse.
Big Brother may never need to issue commands or demand obedience. The greater danger is that governments, corporations or political actors could use AI to shape what we see, what we believe and the choices we make, until their influence becomes almost impossible to recognise.