Sam Altman’s identity verification project, World, has introduced a new tool designed to distinguish artificial intelligence agents from anonymous automated traffic in the financial sector. The company announced on Sunday that its AgentKit is now in beta, offering a solution for businesses to confirm whether an AI system is acting on behalf of a verified human.
AgentKit Beta for Financial Institutions
The new AgentKit feature allows users to delegate World ID Proofs to their AI agents. These agents can then present the proofs to websites or services without exposing the user’s personal data. According to a news release provided to PYMNTS, this setup helps merchants and financial institutions separate legitimate agent activity from the bot traffic that many platforms currently block.
World describes this infrastructure as essential for transactions as AI agents take on more consumer tasks. The company stated that the system provides a way to establish trust in both direct human interactions and those mediated by agents, while ensuring individuals only reveal the specific information required for a given interaction.
Expanding to Video Conferencing
In addition to financial services, World is rolling out World ID for Zoom through the Zoom App Marketplace. This integration addresses a similar trust issue in video meetings, specifically verifying that the person appearing on screen is the intended participant. After a one-time verification using the company’s Orb hardware, users can authenticate themselves before or during a call using a live selfie.
The broader World ID app is now available in more than 150 countries. It allows users to manage proofs of human status, presence, and specific attributes. The company’s core thesis is that these underlying proofs can standardize trust across different digital interactions.
Agentic Commerce and Payment Rules
This launch occurs as the payments industry grapples with the rise of agentic commerce. A recent report by PYMNTS Intelligence and Visa Acceptance Solutions highlighted that current payment stacks are optimized for transaction authorization but lack rules for agent authorization. The report noted that the key question is shifting from whether a transaction should be approved to whether the machine making the decision was authorized to do so.
Research cited in the report found that 56% of consumers are willing to let agents search and compare products, but only 35% would grant access to saved payment credentials. This gap suggests that permission is becoming a new primitive in payments, creating an opportunity for networks and processors to build infrastructure that translates human intent into machine-executable rules.
World ID for banking and finance aims to provide the trust layer needed for this evolving landscape, allowing institutions to verify the human origin of agent-driven actions.
Source: PYMNTS
