At the beginning of 2024, a financial employee in Hong Kong sat in front of his screen to take part in a business video meeting. On the screen, he saw familiar faces—his colleagues and some of the senior executives at his company—and heard their usual voices and recognizable tones. After a brief discussion, he carried out the requested financial transfers without any suspicion. He discovered only later that none of the people who appeared before him had actually been present: they were artificially created digital replicas that used live-video deepfake technology to impersonate several colleagues at once. The incident ended in a loss of approximately $25 million, making it one of the largest publicly reported corporate deepfake fraud cases.
This incident sums up the question of our time
If machines can create a face and voice convincing enough to fool those closest to us, what means of verification do we have left? And how do we balance a technology that promises to transform medicine, education, and work with one that is simultaneously being used to steal money, destroy reputations, and blackmail individuals?
The Promise: A Technology That Has Changed the Pace of Life
It is impossible to understand this phenomenon without acknowledging the scale of what the technology has delivered. Artificial intelligence has spread at an unprecedented pace. The 2026 International AI Safety Report noted that its adoption had outpaced that of the personal computer, with at least 700 million people using leading AI systems every week. This spread is not without reason. The technology helps doctors interpret radiological images and detect certain diseases early; opens access to knowledge for people who do not speak the languages of science through real-time translation; gives people with disabilities communication and work tools that were previously unavailable to them; and saves journalists and researchers long hours of sorting through and summarizing documents.
This technology also has another side that is sometimes overlooked: it is being used to combat fraud itself. The 2026 Pindrop Deepfake Readiness Index found that companies that had deployed specialized detection tools recorded 60% fewer successful deepfake incidents than those relying solely on traditional methods. Yet the same index reveals the troubling side of the picture: only 22% of organizations use these tools.
The Turning Point: When Fraud Became Accessible to Everyone
Impersonating others once required skill, patience, and resources. Today, technology has swept away those barriers. The International AI Safety Report says that many deepfake tools are free or low-cost, require no technical expertise, and can be used anonymously. It adds that underground markets now sell ready-made AI-powered tools that lower the level of skill required to carry out attacks. Fraudsters no longer need to work with a programmer; they can simply purchase a ready-made tool.
The most dangerous aspect of this shift is that the technology has targeted the weakest link in our defenses: trust. We believe our mother’s voice, our manager’s features, and our friend’s tone because human experience has taught us that these signals are not easily forged.
Research has shown that this reassurance is no longer justified. One study cited in the International AI Safety Report found that listeners mistook AI-generated voices for those of real speakers 80% of the time. The threat is not limited to voices. Digital Applied reported in its report, "Deepfake Statistics 2026: Fraud and Detection Data," that deepfakes now account for one in five fraud attempts targeting biometric verification systems.
The New Fraud Map: From Investment Schemes to Family Calls
The methods vary, but they share one principle: exploiting the victim’s trust in a person or entity they know. One of the most lucrative methods is impersonating celebrities and officials to promote fake investments. A Surfshark study concluded that this tactic caused losses of $1.13 billion, or 52% of all recorded losses from deepfake fraud, while another 31% resulted from fraudulent investment opportunities. Even the highest offices are not immune. In May 2026, the Singapore Police Force revealed that a victim had transferred $3.8 million to a fraudster impersonating Prime Minister Lawrence Wong.
On the other side is fraud that hits close to home. The International AI Safety Report warns that fraudsters use voice cloning and deepfakes to impersonate managers and family members and trick victims into transferring their money. Imagine receiving a call in the middle of the night in your child’s voice, pleading for urgent financial help, with every trace of pain and confusion you hear manufactured by algorithms. In such cases, the fraudster is not trying to convince your mind; they are trying to overwhelm your emotions before your mind has a chance to intervene.
Older people pay a particularly high price for this type of deception. The first report in which the U.S. Federal Bureau of Investigation dedicated a category to AI-related fraud found that verified losses totaled $893.3 million, of which $352 million was suffered by people aged 60 and over—approximately 39% of the total.
Figures That Do Not Tell the Whole Story
People tend to seek a final figure for the scale of the losses, but figures in this area vary according to the source and methodology. In April 2026, Surfshark estimated documented global losses at $2.19 billion for the period from January 2019 to March 2026, while later estimates based on the same study put the figure at at least $3.7 billion. The discrepancy stems from the fact that each organization monitors a different stage of the fraud ecosystem and relies on a different database. Some analysts have also warned that there is no single reliable global figure encompassing these losses in the first place.
The situation is further complicated by the fact that most victims remain silent. Analyses suggest that fewer than 5% of voice-cloning victims report their losses. Accordingly, every published figure should be read as a lower bound representing the visible part of the problem, not its full scale.
Beyond Money: Blackmail and the Erosion of Truth
If deepfakes were limited to stealing money, the damage would be confined to the financial sphere. But their impact is deeper. The International AI Safety Report warns that they are being used for defamation, blackmail, fraud, and the production of non-consensual sexually explicit images. It notes that such content is becoming more realistic and harder to detect, and that these fabricated videos disproportionately target women and girls. It found that 96% of deepfake videos circulating online are non-consensual sexually explicit content—a figure showing that the overwhelming majority of this technology is being used against real people without their knowledge.
This type of assault has a particular nature. A victim may be blackmailed with images that were never taken, and may face social stigma for an act they did not commit. Here, the matter becomes more than a cybercrime: it is psychological and social harm whose effects extend beyond the screen.
There is also a more far-reaching danger connected to the very nature of truth. The International AI Safety Report noted that researchers had observed deepfakes undermining the reliability of evidence presented in court. When every video can be called into question, liars benefit from doubt just as honest people suffer from it: someone proven guilty through audio and video can claim that both are fake. In this way, it is the very concept of evidence—not just individual victims—that is harmed.
What Experts Say: Warnings from Within the Industry
Notably, the strongest warnings have not come from opponents of the technology, but from some of its most prominent practitioners. Geoffrey Hinton and Yoshua Bengio—two of the most frequently cited computer scientists and recipients of the Turing Award—signed a statement on the risks of AI alongside scientific and executive leaders at major companies, including OpenAI, Google DeepMind, and Anthropic. Hinton had left Google in 2023 to speak more freely about these risks, a step that gave particular weight to his warnings.
Bengio also chairs the International AI Safety Report, whose second edition was released on February 3, 2026, with contributions from more than 100 experts from over 30 countries. The report concluded that risks are growing faster than the safeguards designed to contain them. Although many protective measures are improving, current risk-management techniques remain vulnerable to failure.
This picture is not confined to laboratories and conferences. A Pindrop survey found that 74% of the security officials surveyed had encountered a suspected deepfake incident during the previous year, and that one-quarter of those affected had suffered losses exceeding $1 million in a single incident. Gartner reported that 62% of organizations had actually experienced a deepfake incident.
The conclusion shared by these diverse parties is that the technology is advancing faster than laws and protective tools, and that relying on technological solutions alone is not enough.
Lines of Defense: From the Individual to Society
If technology alone cannot provide sufficient protection, what can be done? The answer begins with individuals: simple verification habits are more powerful than we imagine. The first is not to trust what we hear or see alone. If a “relative” or “manager” asks you for money urgently, end the call and call them back using their known number. Families can agree on a password to be used only in emergencies; it can expose a cloned voice within seconds. We should be wary of urgency and alarmism, because psychological pressure is a fraudster’s primary tool, and no financial decision is worth making under threat. It is also wise to limit the voice recordings, images, and personal information we post publicly, as these are the raw materials used to create fake replicas. If we fall victim to blackmail, we should not pay; we should preserve the evidence and report the matter to the relevant authorities immediately.
At the institutional level, the figures show that investing in detection tools and training employees in protocols for verifying unusual financial requests is not a luxury. At the societal level, the media and public-awareness platforms have a pivotal role to play by explaining these risks in accessible terms and encouraging a culture of verification before belief and sharing.
A Tool That Reflects the Person Using It
Artificial intelligence is neither wholly evil nor wholly good. It is a powerful tool that amplifies the capabilities of whoever wields it—whether a doctor making a diagnosis or a fraudster deceiving someone. Deepfakes have moved from a technological curiosity to a fixed item in organized crime budgets. Their effects are not limited to major powers: the same tools are available in our languages and dialects, and Arab readers can be targeted just like anyone else.
That is why the most important rule in the age of deepfakes is to maintain a distance between feeling and belief, and to make verification a habit before it becomes a necessity. The voice of someone we love was once our final sign of safety; today, we must learn to listen to it with caution.
