
The rise of deepfakes and synthetic identity fraud
Whether it’s approving a payment on a video call, responding to a voice note from a colleague, or joining a meeting with senior leadership, modern business relies heavily on digital interaction. Voice, video, and real-time communication have become central to how organisations operate and make decisions.
However, in a world increasingly shaped by artificial intelligence, do you truly know who you’re speaking to?
For years, we’ve instinctively trusted human signals such as a familiar voice, a recognisable face, or a known name on a screen. These cues have traditionally been sufficient to establish confidence in the person on the other end of a conversation.
But today, we can no longer rely on these cues to confirm that a person is who they claim to be, or even that they are a real human being.
The landscape
Identity fraud has long been a significant risk to organisations. In fact, in the UK, 2025 saw more than 1,200 fraud cases each day, with more than half of them linked to identity fraud.
Historically, attackers focused on stealing login credentials, hijacking accounts, or impersonating known individuals using compromised data. In 2026, however, the threat has evolved.
Advances in artificial intelligence have made it possible to generate highly convincing voice clones, realistic deepfake videos, and entirely synthetic personas. These capabilities allow fraudsters to bypass traditional verification processes and impersonate individuals with a level of accuracy that was previously unimaginable.
According to the 2026 Thales Data Threat Report, 59% of organisations have already experienced deepfake-related attacks, demonstrating how widespread and impactful this threat has become.
At the same time, the Cloudflare 2026 Threat Report highlights how artificial intelligence has dramatically lowered the barrier to entry for cybercriminals, enabling even less sophisticated actors to carry out highly effective and scalable attacks.
Why deepfake fraud is effective
We trust human signals instinctively. In a physical environment, identity is rarely questioned because it is easy to recognise who is present. In digital environments, however, organisations rely on voice, video, and display names to replicate that sense of trust. If you joined a Teams call with a senior leader who looked and sounded like they usually do, would you even question it?
Unfortunately, these typical trust signals can now be convincingly faked. Voice cloning technology can mimic tone and speech patterns, while deepfake video can replicate facial expressions and movements with increasing realism.
While much of the public attention around AI fraud has focused on deepfake video, synthetic voice may represent the greater threat. Unlike facial imagery, which is largely passive, a voice is an active communication channel. It is the voice that delivers instructions, authorises transactions, and influences decisions. A fraudster does not need to simply look like a trusted executive; they need only sound like one and say, "Please transfer the funds," or "I need you to approve this payment immediately."
The risk is even greater in voice-only environments such as telephone banking, contact centres, and customer service operations, where there is no visual cue available at all. In these situations, a cloned voice may be the only identity signal an employee has available to assess legitimacy.
For example, a contact centre agent is typically measured on customer experience metrics such as customer satisfaction and Net Promoter Score (NPS). In practice, very few agents are likely to challenge a caller by saying, "I cannot complete this request because I think you might be synthetic." Even if something sounds slightly unusual, the pressure to provide a seamless customer experience can make these situations extremely difficult to navigate.
As a result, the cues we rely on to confirm identity are no longer reliable indicators of authenticity. The Cloudflare Threat Report explains that threat actors are actively using deepfake audio and video to impersonate individuals during live communications.
The real-world impact
The consequences of deepfake and synthetic identity fraud are already being seen across industries.
In early 2024, an employee at UK engineering firm Arup approved a £20 million transfer after joining what seemed to be a legitimate video call with senior leadership. In reality, the entire interaction had been manipulated using AI-generated deepfakes, with attackers convincingly impersonating multiple executives.
In another case, fraudsters targeted global advertising giant WPP by creating a WhatsApp account using a publicly available image of CEO Mark Read. They then arranged a Microsoft Teams call, using a cloned voice and repurposed video footage to convincingly impersonate the company's top executives. During the call, the attackers tried to persuade an employee to create a new business entity that could be used to siphon funds and sensitive information.
Although the WPP attack was ultimately unsuccessful, it prompted a company-wide warning. The CEO emphasised the growing sophistication of these tactics, noting that employees must remain vigilant to threats that go beyond traditional phishing and now exploit video, voice, and real-time interaction.
Why traditional verification is no longer enough
Traditional biometric authentication methods that only verify identity at login, for example, face-ID, fingerprint scanning or voice recognition are no longer designed to protect against AI-powered threats. This approach creates a clear vulnerability. If that one signal is compromised, for example through the use of a deepfake or a synthetic or cloned voice, the entire security system can be bypassed.
In addition, one-time identity verification methods such as voice recognition assume that, once a user passes the initial check, they remain the same individual throughout the interaction. Without continuous authentication, however, organisations cannot be certain that the person who logged in is still the one behind the screen or on the other end of the phone.
Instead of relying on a single identifier, continuous fused biometric verification technology, like FARx, combines multiple biometric modalities, including facial, voice, and speech pattern recognition. By layering these signals, organisations can ensure that access is granted only when the person is genuinely who they say they are and that individual remains the same verified human throughout the entire interaction.
This makes it significantly harder for fraudsters to impersonate an individual using deepfakes, voice clones, or other forms of synthetic identity fraud - delivering critical identity security, compliance assurance, and data privacy protection.
To learn more about how fused voice-face biometrics can be used to reliably confirm identity, prevent fraud, and enhance security, get in touch with our expert team here.
Get in touch with our expert team
Discover how FARx can strengthen identity assurance across your organisation
Get in Touch






