World ID Targets AI Deepfakes in Finance With Human Verification for Zoom and AI Agents
World Network is bringing its World ID technology into banking and finance as increasingly convincing artificial intelligence deepfakes raise concerns about identity fraud and financial security. The initiative introduces tools designed to help institutions verify that people participating in sensitive interactions are genuine humans and that AI agents acting on their behalf have authorized credentials. World says the technology uses privacy-preserving cryptographic proofs rather than relying solely on conventional identity checks.
The announcement highlights a widely discussed example of deepfake-enabled fraud. In 2024, an employee at engineering firm Arup reportedly transferred $25.6 million after joining a video call in which the other participants were deepfake impersonations. The incident demonstrated how convincing audiovisual deception can bypass human judgment, particularly when employees believe they are communicating with senior executives or trusted colleagues.
World argues that the risk is growing as AI systems become more capable of imitating people and autonomous agents take on additional financial tasks. In this environment, recognizing a familiar face or hearing a familiar voice may no longer provide sufficient assurance that a person is genuine. Financial institutions need stronger ways to verify identities, permissions and the origin of instructions before authorizing sensitive actions.
How World ID Is Designed to Verify People and AI Agents
World ID for Zoom is intended to let meeting participants confirm that they are the same human who completed verification through World’s Orb-based process and is associated with their account. The goal is to provide an additional layer of assurance during meetings where participants may discuss transactions, approve decisions or share confidential information.
The approach is different from simply checking whether a video appears realistic. Deepfake technology can reproduce faces, voices and expressions, making visual inspection an unreliable safeguard in some circumstances. Cryptographic proofs can help verify a credential or identity-related claim without requiring every participant to receive the person’s underlying personal information. The actual level of protection, however, depends on implementation, account security and how organizations use the verification results.
World is also introducing World ID for agents, which allows AI agents to carry proofs delegated by people. This is relevant as software agents increasingly interact with online services, retrieve information and potentially initiate financial actions. A delegated proof could help a service establish that an agent is acting with a human’s authorization, rather than treating every automated request as an independent and unverified identity.
Such credentials do not automatically make an AI agent safe. Institutions would still need to define the scope of delegated permissions, determine which actions require additional approval and revoke access when authorization expires or circumstances change. Verification can help establish who or what is presenting a credential, but it does not guarantee that a transaction is legitimate or that an agent will make the correct decision.
Financial Institutions Explore New Trust Infrastructure
World identified Bain Capital, OpenExchange, Outtake, Pantera Capital and VanEck Funds as initial design partners for the initiative. Their participation signals interest in testing identity and verification tools for financial interactions, although it does not establish that every organization has fully deployed the technology across its operations.
For banks, investment firms and financial service providers, the potential applications extend beyond video meetings. Identity proofs and delegated credentials could support account access, approvals, communications with counterparties and workflows involving automated software. Any deployment would need to fit existing compliance requirements, cybersecurity controls and internal risk-management procedures.
Privacy will also be an important consideration. World describes its system as based on privacy-preserving cryptographic proofs, but institutions and users will still need clear information about what data is collected, how credentials are issued and revoked, and what happens when verification fails. The reliability of the verification process and the security of the systems connected to it will matter as much as the underlying cryptography.
The broader significance is that financial security is expanding beyond protecting passwords and preventing unauthorized account access. Institutions must also determine whether instructions come from the right person and whether an AI agent has permission to act. World ID’s move into finance addresses part of that challenge, but adoption and real-world testing will determine its effectiveness. As deepfakes become more convincing and automated financial activity increases, identity verification may become a more important layer of financial infrastructure rather than an optional meeting feature.
About The Author
