A face can be fabricated. A voice can be cloned. An event can be invented. But who decides what remains true?
There was a time when seeing was believing. A photograph carried the weight of evidence, a recorded voice could establish a fact, and a video could preserve a moment that had already happened. The world was not necessarily more truthful then, but evidence had a certain authority. What was seen, heard and documented could be questioned, challenged and investigated, but it could not be casually manufactured at scale. That confidence is now beginning to disappear. A face can be digitally created, a voice can be replicated, a person can be made to appear to say words they never spoke, and an event that never happened can be presented with the visual authority of reality. The technology is advancing at a remarkable speed, but our ability as a society to distinguish reality from fabrication is struggling to keep pace. The deeper danger of the deepfake era, therefore, is not simply that people may believe a lie. It is that people may eventually stop believing the truth.
Deepfakes are often described as a technological problem, but that description does not go far enough. The real challenge is much larger because it strikes at the relationship between evidence and public trust. A lie once needed to be convincing. Today, it increasingly needs only to be plausible. Once people understand that almost anything can be fabricated, genuine evidence can also be dismissed as artificial. A politician can deny an authentic recording by claiming it was generated by artificial intelligence. A public figure can question a genuine photograph. An individual accused of wrongdoing can dismiss incriminating material as manipulated. The defense no longer necessarily requires proving that something is false. It can simply create enough uncertainty for people to stop believing that it is true. This is perhaps the most sophisticated danger of deepfakes: they do not merely manufacture falsehood; they manufacture doubt.
That doubt becomes particularly dangerous when it enters politics. Democracy depends on citizens having access to information from which they can make judgments. People can disagree about policies, governments and political ideologies, but there must still be some shared understanding of what actually happened. When that foundation begins to collapse, democratic debate itself becomes vulnerable. A fabricated video of a political leader, a manipulated recording, a synthetic announcement or an artificial image can spread through social media within minutes, reaching millions before journalists, authorities or fact-checkers have established its authenticity. By the time the truth emerges, the falsehood may already have shaped public opinion. The correction may be accurate, but accuracy does not guarantee reach. A lie can travel at the speed of outrage; truth often arrives carrying the burden of verification.
This is why the response to deepfakes cannot be reduced to simply creating more laws. Governments have a legitimate responsibility to protect citizens from malicious impersonation, fraud, election manipulation and deliberate deception. Technology platforms also cannot continue to treat the consequences of content distribution as somebody else's problem. But there is a delicate line that must not be crossed. The fight against misinformation cannot become an excuse for governments to acquire unchecked authority over what citizens are allowed to believe or discuss. The cure for manufactured truth cannot be state-controlled truth. No government, political party, technology company or media organization should be permitted to become the ultimate owner of reality. Truth must remain accountable to evidence, independent verification and public scrutiny.
This places an enormous responsibility on journalism. The traditional newsroom instinct has often been to ask whether a story can be published. In the age of deepfakes, the more important question is whether the story can be proven. The difference is fundamental. A journalist cannot treat virality as verification, popularity as authenticity or emotional impact as evidence. A million views do not make a video real. A familiar voice does not prove that the person actually spoke those words. A photograph does not become evidence merely because thousands of people have shared it. Journalism must therefore become more rigorous precisely when the digital environment demands greater speed. The pressure to publish first cannot be allowed to become more important than the responsibility to publish correctly. The newsroom that wins the race to spread a falsehood may gain temporary attention, but the newsroom that takes the time to establish the truth earns something far more valuable: trust.
The responsibility, however, does not end at the newsroom. The public has become part of the information system itself. Every person carrying a smartphone can now become a distributor of information, whether verified or false. A forwarded video can reach hundreds of people without passing through a single editorial desk. A screenshot can become a claim. A fabricated recording can become a conversation. A rumour can become a perceived fact simply because enough people have repeated it. This makes digital literacy not merely a technical skill but a civic responsibility. Before sharing something shocking, people must ask where it came from, whether independent sources have confirmed it and whether the material is designed to provoke an immediate emotional reaction. The more sensational the claim, the greater the need for caution. In the age of deepfakes, pausing before sharing is no longer simply good online behaviour. It is part of protecting the information environment in which democracy operates.
There is also a deeply personal dimension to this crisis that cannot be ignored. A fabricated video does not need to influence an election or destabilize a government to cause serious harm. It can target an ordinary person whose face, voice or identity is manipulated and distributed without consent. A student, teacher, professional, public servant or private citizen can become the subject of a false narrative that spreads faster than any possible defense. Reputation can take years to build and minutes to damage. Once manipulated content enters the internet, removing the original does not necessarily remove the copies, screenshots and reuploads that have already escaped into the wider digital world. The principle that a person should not be condemned without evidence becomes increasingly difficult to protect when accusation can travel instantly and correction struggles to catch up.
Yet blaming artificial intelligence itself would be an easy and incomplete response. Technology is not inherently the enemy. The same tools capable of generating convincing synthetic media can also contribute to education, research, accessibility, entertainment and journalism. The problem lies in how those capabilities are used and whether institutions are prepared to establish meaningful boundaries around them. The technology will continue to improve, and attempts to stop every malicious use will inevitably face limitations. The more important task is to build systems and habits that make deception harder to believe and easier to expose.
The question, then, is not simply how we detect deepfakes. It is who we trust when detection itself becomes contested. Governments must protect citizens without becoming arbiters of acceptable reality. Technology companies must accept responsibility for the environments they create without becoming invisible governors of public discourse. Journalists must maintain independence while strengthening verification. Citizens must remain sceptical without becoming cynical. There is a crucial difference between questioning evidence and rejecting evidence altogether. The first is essential to a healthy democracy; the second can make democracy incapable of establishing anything at all.
The deepfake era will not disappear. Synthetic voices will become more convincing, artificial images will become harder to distinguish from photographs and manipulated videos will become increasingly sophisticated. Detection technologies will evolve alongside them, but this will not be a battle that technology alone can win. The deeper defence must be a culture of verification: newsrooms that value accuracy over virality, institutions that accept accountability, platforms that recognize responsibility and citizens who understand that sharing information is itself an act with consequences. We cannot prevent every lie from being created, but we can build a society that does not surrender its judgment every time a convincing image appears on a screen.
The most dangerous future is not one in which everyone believes every deepfake. It is one in which people become so accustomed to manipulation that they believe nothing can be trusted. If every photograph can be dismissed as artificial, every recording questioned as synthetic and every video rejected as fabrication, accountability itself becomes weaker.
Power no longer needs to prove its innocence or explain its actions; it only needs to create sufficient doubt about the evidence. That is why the battle over deepfakes is ultimately not a battle over artificial intelligence. It is a battle over whether facts will continue to have authority in public life. Technology may manufacture faces, voices and events, but it must never be allowed to manufacture reality itself. Truth cannot belong to whoever has the largest platform, the loudest voice or the most sophisticated algorithm. Truth cannot be owned. It can only be verified, defended and held accountable to evidence.