In early 2024, an employee at global engineering firm Arup joined a video call that appeared to include the company’s chief financial officer and several colleagues.
The employee had initially suspected the request might be a phishing attempt. The call removed that suspicion, and the employee transferred roughly $25 million.
The attackers had not cracked a password or exploited a software vulnerability. They had built convincing digital replicas of people the employee recognized. Standard fraud controls were not designed to detect the difference between a genuine video call and a fabricated one.
Financial institutions are now weighing what verification should look like when a voice, face, or digital identity can be replicated with AI tools.
World, a digital identity company co-founded by Sam Altman, brought a suite of products for banking and finance to Money20/20 on Oct. 18, positioning human verification as part of the answer.
Also read: Michael Burry just made another bold move on Wall Street
Where AI-driven impersonation finds its opening in banking
Most financial security systems check whether someone holds valid credentials. A correct password, a registered device, or a one-time code confirms that someone controls an account. Those checks do not independently confirm that the person behind the credentials is who they claim to be.
The gap persists because credentials are proxies. A password, a registered device or a one-time code confirms that someone controls an account. It cannot confirm whether the person using those credentials is the account’s legitimate owner.
The gap creates two distinct exposures. For individual customers, a fraudster who takes over an account can initiate transfers that the system processes as legitimate.
For enterprises, stolen credentials can be used to access confidential financial data or authorize payments without triggering standard controls. The 2024 Arup transfer cleared existing security systems without flagging an alert.
Deloitte estimated that AI-enabled fraud losses in the United States could reach $40 billion by 2027, up from roughly $12.3 billion in 2023. The estimate reflects how AI tools have lowered the cost and technical barrier to producing convincing impersonations.
SDI Productions / Getty Images
The push to add a human layer on top of existing security
Financial institutions have begun exploring verification approaches that go beyond confirming credentials and check whether a specific, verified human is present. The shift reflects a broader recognition that the credential layer was not built to catch impersonation at the level AI tools now make possible.
At Money20/20 on Oct. 18, World co-founder Alex Blania announced World ID for banking and finance, a suite that combines an identity wallet for storing verified human proofs, a video verification tool for confirming whether a meeting participant matches a previously enrolled identity, and a framework for delegating verified proofs to AI agents acting on a person’s behalf.
More than 18 million people hold a Proof of Human credential through the World network.
Ajay Patel, head of World ID at Tools for Humanity, told TheStreet that the standard for high-stakes interactions needed to shift. “Knowing that an account is authenticated is important; having confidence in the human behind it should become part of the foundation, rather than an exception reserved for when something looks suspicious,” he said.
World is not the only company working in this space. Digital identity verification has drawn increased investment as financial institutions contend with fraud that existing controls were not designed to catch.
Bain Capital, VanEck Funds, and OpenExchange are among the organizations exploring how proof-of-human technology applies to banking use cases.
Why proving someone is human is only part of the answer
Yuannan Yang, a security engineer at CertiK, said the threat has grown beyond what identity verification alone addresses. “The biggest threat goes beyond deepfakes or voice cloning,” Yang told TheStreet. “What’s more concerning is how AI is making social engineering more convincing, personalized, and scalable.”
The 2024 Arup case illustrated that point. The employee who authorized the $25 million transfer had been alert to phishing before joining the call.
What overcame that judgment was a fabricated social environment. The video appeared to include colleagues the employee recognized. Confirming that the participants were genuine, verified identities would have caught the fraud. Yet confirming only that they were live humans would not have been enough on its own.
More Wall Street:
- Nvidia and Micron are about to lord their successes all over Wall Street
- Nvidia, Palantir, and Alphabet send waves through Wall Street
- Wall Street elites will pocket windfall from record $46 billion profits
Proof-of-human verification addresses one layer of a larger problem, he said. “Even if a bank successfully verifies that a real person is initiating a transaction, that person could still be acting under fraudulent instructions or being manipulated by a scammer.”
A full response to AI-driven fraud requires combining identity verification, device authentication, transaction monitoring, and independent authorization within a risk-based framework, Yang said.
The goal is not only to establish that a real person is present but to confirm the person’s identity and that the action they are authorizing is legitimate.
What banks will need to see before adopting at scale
Banks evaluating proof-of-human technology will consider whether the security benefit justifies the integration work. The approach needs to fit into existing identity, authorization, and compliance systems without requiring institutions to replace them or build new centralized biometric databases.
Banks will also weigh the user experience. Verification steps that slow legitimate customers down can reduce adoption of digital services. Any proof-of-human check will need to deliver stronger assurance without adding meaningful friction beyond what users already expect from standard authentication.
Regulatory attention to the problem has been growing. In October 2025 remarks reported by American Banker, Federal Reserve Governor Michael Barr cited the rapid growth of AI-driven impersonation as a threat to banks of all sizes.
The sector has not previously required proof-of-human checks at scale, and no institution has yet produced performance data from a broad deployment.
Adoption will depend on whether early pilots demonstrate that the technology reduces fraud losses without adding friction that legitimate customers notice.
Related: Nvidia, Palantir, and Alphabet send waves through Wall Street
