AI is changing the economics of fraud. Activities that once required significant human effort can now be automated, allowing fraudsters to operate at greater speed and scale.
This creates a new challenge for online safety. For years, identity verification has focused on establishing whether someone is who they claim to be. As AI evolves, organizations must also consider whether the interaction they are assessing is being driven by an automated adversary.
Telling humans from machines is only part of the answer. The next stage of identity verification will require AI systems that can defend against other AI systems.
From digitization to AI vs. AI
The evolution of identity verification can be understood in four generations.
The first brought manual identity checks online, digitizing processes that had traditionally taken place in person, so organizations could verify identities remotely.
The second generation automated those processes, improving the speed and scalability of digital verification. The third introduced an AI-first approach, using artificial intelligence as a core part of how identity is verified. This is where the industry is today.
The fourth generation represents the next step: AI defending against AI.
It is an important distinction. The technology used to commit fraud is evolving alongside the systems designed to prevent it. Fraudsters can now use AI to target automated processes, exploiting systems that were not designed for the speed and sophistication of today’s threats.
Identity verification must evolve accordingly. Instead of relying on approaches developed for a different threat landscape, systems will need to identify and respond to suspicious activity in milliseconds.
When AI becomes the attacker
AI is already changing how convincing fraudulent identities can be created. Synthetic faces, manipulated video, cloned voices and fabricated documents can make it harder to determine whether an identity is genuine.
But the challenge extends beyond the creation of fake content. AI can also be used to automate attacks against the verification process itself.
Tasks that once required a person to test a system, assess its responses and adjust an approach are now being automated. This allows fraudsters to attempt attacks at greater scale, while adapting their methods to the systems they encounter.
The result is a shift in the nature of the threat. Identity verification is no longer simply a question of whether a document looks genuine or a face matches a photograph. Organizations must also consider whether the process they are assessing is being manipulated by an automated adversary.
This is why AI-powered fraud requires an AI-powered response. Verification systems need to keep pace with attacks that can operate continuously and at machine speed. Approaches built around slower, more predictable threats risk falling behind as automation becomes more sophisticated.
Why speed is now essential
As AI makes fraud more scalable, the time taken to identify and respond to suspicious activity matters more than ever.
A system that cannot respond quickly enough may give an automated attacker more opportunities to exploit weaknesses. Defenses must therefore be capable of assessing risk and responding in milliseconds, rather than relying solely on processes designed for an earlier era of digital identity.
This does not mean that every interaction should be treated as fraudulent or that automation itself is inherently suspicious. AI will continue to play a legitimate role in digital services. The challenge is to distinguish legitimate activity from attempts to exploit the systems that enable it.
That requires identity verification to remain effective as both genuine users and fraudsters adopt new technologies. The objective is not simply to identify individual signs of manipulation, but to ensure that verification can withstand attacks that are increasingly automated.
Onboarding is where this matters most. It is the point at which businesses establish whether a new user can be trusted. If fraud gets through at this stage, it creates risk for every service and accounts that follows.
The next era of digital trust
The implications extend beyond any single verification method. Faces, documents and other identity signals remain important, but the systems assessing them must evolve alongside the threats they encounter.
Organizations need to recognize that the next generation of identity verification will be defined by its ability to defend against AI-driven attacks, not simply by its ability to use AI to verify identities.
