By Hol Thomas-Wrightson, 16 November 2026
Criminals have always used new technology to their advantage, and the AI revolution seems to be offering them new opportunities every day.
Deepfakes and generative AI tools for information and imagery manipulation are just the tip of the iceberg, making it easier for bad actors to fool both individuals and organisations.
It is integral that organisations arm themselves with the right combination of professional knowledge and technological analysis to defend both their business and customers from harm.
Pekka Dare, President of ICA, led a panel on a deep dive into the subject of synthetic identities. He discussed the pressing topic with panellists Rob O'Farrell, Chief Technology Officer and co-founder, ID-Pal, Josie Welland, Senior Managing Associate, Sidley Austin, and Andrei Skorobogatov, Director of Policy for the Global Anti-Scam Alliance (GASA).
They explored the challenges organisations are facing in tackling these developing threats, and what they can do to really know their customers.
Theft vs fabrication
With new terms coming into the compliance lexicon, O’Farrell offered a distinction between the newer issue of synthetic identities verses the more traditional identity fraud framework. ‘Essentially, it comes down to whether you’re taking one person's identity and trying to impersonate them, or taking some real data and some false data to try and create a completely synthetic identity, a new identity that doesn't exist,’ he explained.
Whereas identity fraud involves stealing an individual’s information and using it as found, synthetic identities are created by taking smaller details from multiple sources and using them as building blocks to manufacture a convincing new persona in order to trick both people and technology. ‘Fraudsters have stopped forging documents,’ Welland added. ‘They have started building people.’
In the ongoing identification and verification arms race, generative AI has become a key weapon in the criminal arsenal. O’Farrell shared a video demonstrating the wide variety of both criminal and (at least currently) legal applications. These range from using diffusion modelling to create highly convincing ID documents, or real time video call overlays built from just a single image, to video adapted specifically to fool liveness checks.
O’Farrell emphasised that the issue is not just defending against individual bad actors, but criminal groups acting at a terrifyingly industrial scale. ‘The people that we're up against, they have full business plans, and org charts, and targets for their staff.’
Knowing the risks
It’s not just organisations that are facing attacks. Skorobogatov stated that scams are one of the biggest crime types globally, with synthetic identities enabling more convincing interpersonal scams such as pig butchering, romance scams and investment fraud.
‘Fraud and scams – they’re the crime of deception – and some sort of stolen identity, or fake identity, or synthetic identity, is fundamental to the execution of that,’ he said. AI automation also provides the opportunity for a fraudster to launch attacks at a far larger scale than they would be able to manually.
Besides the obvious fraud and money laundering risks, Welland brought up the aspect that compliance professionals will be wondering about: regulatory risk. ‘A single incident of a Synthetic ID passing through is not going to be a breach,’ she reassured, ‘But it’s when it becomes a pattern that it becomes a problem.’ UK regulators will focus on three main factors in this space: the potential or actual exposure to loss, handling the proceeds of crime, and the potential of money laundering. ‘Loss is the starting point, the visible harm, but exposure is much more about how access was used and the controls put in place,’ Welland added.
Divided we fall
Information sharing plays a key role in fighting these types of fraud. As Dare put it, many organisations treat fraud and anti money laundering functions as separate, disparate sectors that work in silos. He noted: ‘They are intrinsically linked, right? If you have a successful fraud, by definition, you've got money laundering.’
But the issue expands beyond the bounds of any single organisation. While individual firms may be able to build an understanding of the threat signals that are recognised in their own cases, if they are not able to share and compare findings across organisations and industries, they’re working at a huge disadvantage. ‘You need to be able to share those signals and to be able to create the full picture,’ Skorobogatov said, ‘and off the back of the picture, it is a lot easier to identify a potential threat actor.’
‘You are not alone,’ he added. ‘It takes a network to defeat a network.’
Reality check
When it comes to building a process for checking, O’Farrell recommends what he calls ‘the chain of trust’.
‘If you are trying to verify an individual,’ he explained, ‘but you decide to skip a step anywhere along the way in the checks, all subsequent checks after the point you skipped are no longer effective.’
He recommends the following checks to help reduce risk of fake information getting through.
Check that there is a:
- real camera being used – make sure that image or video evidence has come from the individual’s own camera, rather than an injection attack where artificially created content is added into the process
- real ID – where a photocopy or print out of identification is provided, use presentation attack detection to make sure that it’s true to an original, not a doctored screen recapture or edited document
- real person that matches the ID – next you need to make sure that the person you’re interacting with actually matches the information on identification, that they are not using a mask or real time overlays or screen recapture, and that you’re using appropriate facial recognition.
It’s also important for organisations that use external vendors, that they ensure those vendors are looking at the right things with the appropriate level of scrutiny.
‘Don't be thinking about how complex every single detail of what people are doing at every step is,’ O’Farrell encourages. ‘Just think about the steps in the chain, and make sure that you're dealing with the true owner of an identity, not just a real identity.’
The full webinar Synthetic identities: Do you really know your customer? is available to ICA members via our Learning Hub.
For more information and to sign up to our upcoming ICA webinars, visit our events page.
You may also be interested in: