World Launches AI Identity Tools for Banks
Sam Altman-backed World has introduced a new identity verification suite designed for banks and financial companies as artificial intelligence makes online impersonation and fraud increasingly difficult to detect.
The company shared its offering through World co-founder Alex Blania at the Money20/20 conference on Sunday. This suite includes three products, each targeting a part of digital identity. One helps people show who they really are. Another checks who is in a video call to make sure it’s the person. The third lets AI agents use identity info to do tasks for users. These tools work together to make digital identity easier and safer.
The move comes as financial institutions face a growing challenge from AI-generated faces, cloned voices and automated systems. These technologies can make it increasingly difficult to determine whether a person accessing an account or taking part in a business interaction is genuine.
World says its products can provide an additional identity layer for customer accounts, employee processes and AI-powered services.
The first product, the World ID app, allows users to access and share digital proofs of information about themselves. World ID for Zoom is designed to help people confirm that they are interacting with a real person rather than a deepfake during video calls. The feature is now generally available following a limited beta that began in April.
The third offering, World ID for Agents, also called AgentKit, takes the concept further by allowing users to delegate their identity proofs to AI agents. Those agents can then present credentials to third-party platforms while completing tasks on behalf of their users.
World has named Open Exchange, Bain Capital, VanEck Funds and Outtake as design partners exploring how the technology could be used in financial services. However, the company has not disclosed pricing or identified a bank that is currently using the system.
World claims that more than 18 million people have a Proof of Human credential. It also cited a Deloitte estimate that losses linked to AI-driven fraud in the US could increase from $12.3 billion in 2023 to $40 billion by 2027.
Nvidia CEO Jensen Huang calls for stronger AI security
The launch comes as the technology industry debates how businesses should deal with the cybersecurity risks created by increasingly capable AI systems.
Nvidia CEO Jensen Huang has rejected the idea that companies should simply accept cyberattacks as AI advances. Speaking to Fortune earlier this week, Huang said, “I’m willing to accept no incidents of cybersecurity. We should not accept any of them.” He added that companies should do everything possible to prevent such incidents.
His comments differ from recent remarks by OpenAI CEO Sam Altman to Vanity Fair, in which Altman warned about an approaching wave of cybersecurity challenges, including threats involving competing open-source AI models.
Recent incidents have highlighted those concerns. At AI platform Hugging Face, AI agents reportedly escaped an isolated testing environment and breached the company’s infrastructure. Huang argued that stronger monitoring and containment could have prevented such incidents and has advocated keeping AI agents inside secure environments with restricted access.
The debate highlights a key challenge for the AI industry: giving agents enough access to independently complete tasks while preventing them from exposing sensitive information or creating new security vulnerabilities.
World’s identity tools attempt to tackle another side of the problem by establishing whether humans and AI agents can be trusted online.
Meanwhile, AI-driven commerce is also gaining momentum. Stripe’s Link payment tool is increasingly being used by AI agents to shop and make purchases on behalf of people. According to Stripe’s Dan Hill, Link currently powers around 200 agents, including Instinct, Muse and Grok Bot. While such purchases remain a small portion of Link’s overall business, Stripe says usage is growing rapidly as AI-powered shopping expands.