The global AI agent market, projected to reach $236 billion by 2034, is rapidly expanding into core enterprise functions. For businesses, this means a growing share of customers won’t be humans. They’ll be agents acting on behalf of individuals, interacting with other agents representing sellers, logistics providers and payment processors. Most of the commercial supply chain will eventually be agent-to-agent.
When a human clicks “buy,” identity is reasonably well-understood. When an AI agent does it on their behalf, you have a chain of custody problem: Is this agent authorized? By whom? For what? With what limits? And who is liable if something goes wrong? Today there are no clear answers.
As more AI agents become capable of acting on behalf of people by initiating transactions, accessing services, and making decisions autonomously, organizations can no longer rely on identity verification as a one-time onboarding event. Businesses need a persistent identity layer that continuously verifies people, businesses, and AI agents across every interaction.
That is where Prove, which recently joined the World Economic Forum’s unicorn community, comes in. The New York City-based digital identity scale-up, which has thousands of enterprise customers, including 19 out of the U.S.’s top 20 banks, has developed what it calls the Prove Identity Platform.
Most identity systems today are point solutions — verify once at onboarding, authenticate at login, monitor fraud separately. Prove’s positioning is to shift identity from a one-time event to a persistent trust layer. It authenticates individuals in real time using years of behavioral, device, and telemetry data – a continuously updated picture of who a person is, how they transact, and whether they authorized a specific action. Its agentic suite – which is currently in beta – embeds cryptographically signed consent directly into an identity token, so it travels with every agent action — a single token call that delivers proof of humanity, proof of consent, and proof of authorization.
Think of it as an “agent passport” that ties together who the user is, what they approved, which payment credential is allowed, and which software agent can act. That’s the missing trust step when a bot, not a person, clicks buy, says Mary Ann Miller, Prove’s Fraud & Cybercrime Executive Advisor and VP of Client Experience.“Our mission is enablement of future innovation and establishing trust,” she said in an interview with The Innovator. “Identity is part of it.”
Prove’s platform is built on open protocols already integrated across AI frameworks including OpenAI, Anthropic, and Salesforce, and draws on 12 years of authenticated identity history across 2,000+ enterprise customers.
Its goal is to embed its agent passport deep into the framework of large card issuers as part of an identity enablement layer that would be built into the agentic flow, says Miller.
State-of-Play
AI agents are already initiating real transactions on behalf of real people. OpenAI and Stripe launched the Agentic Commerce Protocol in September 2025. Visa has named Anthropic, OpenAI, and Perplexity as agentic commerce partners.
The idea is to get ahead of potential problems: Analysts project that one in four enterprise breaches by 2028 could stem from AI-agent exploitation.
Prove says its approach is resilient because it is rooted in signals that cannot be deepfaked, phished, or stolen: longitudinal digital history accrued over time, not a document or a face captured in a single moment.
Other companies trying to tackle this space include U.S. companies Ping Identity and Socure.
Ping Identity, a Denver, Colorado-based software company that specializes in identity management solutions, announced an AI framework in late 2025 focused on closing the “trust gap” created by AI agents — issuing unique credentials to agents, distinguishing legitimate from malicious activity, and keeping humans in control with approval workflows.
Socure, another member of the Forum Unicorn community, has developed a platform that it says unifies identity verification, fraud prevention, compliance, credit, and risk decisioning into a single intelligence layer that continuously learns from global signals and adapts in real time as risk patterns evolve. Unlike point solutions that tackle fraud in isolation, RiskOS aims to deliver an integrated system of record for trust, purpose-built for the era of AI-driven threats, with more than 200 third-party integrations that provide a holistic view of risk.
Know Your Agent
In an article on the Forum’s website Socure called for a Know Your Agent (KYA) framework that would work alongside traditional Know Your Customer (KYC) requirements.
A functional KYA framework would need to zero in on four capabilities: establishing who and what the agent is; confirming what it’s permitted to do and for whom; maintaining clear accountability for every action it takes; and continuously monitoring its behavior to ensure it remains within approved parameters, CEO Johnny Ayers said in the posting.
The next decade will determine which version of the agent economy we inhabit, he says. If action is taken now “we could unlock frictionless, cross-border digital commerce where agents transact with high accountability and minimal friction,” he says. Agentic AI could deliver $3 trillion in corporate productivity gains globally over the next decade, expanding access for small businesses and enabling entirely new layers of economic activity.”
But “if we fail bad actors will deploy malicious agents capable of large-scale impersonation and automated fraud,” he predicts. “Trust wouldn’t fade gradually; it would collapse, triggering regulatory overreach that stifles innovation or fragmenting the Internet into isolated, heavily policed walled gardens.”
“We’re not waiting for standards to emerge,” says Prove’s Miller. “We’re building them. With 12 years of authenticated identity history, deep integrations across the world’s leading AI frameworks, and an agent passport that travels with every transaction, Prove is defining what identity means in the agentic era. No one else has the longitudinal data, the enterprise relationships, or the open-protocol architecture to do this at scale. We’re writing the rules for the agentic age.”
