By: Frances Zelazny, General Manager, New Market Initiatives at Prove
Image: Image: Lilartsy - Unsplash
The Five Eyes intelligence alliance between the U.S., U.K., Canada, Australia, and New Zealand recently issued a rare joint statement: the potential for devastating, AI-powered cyberattacks is months (not years) away. The Five Eyes statement comes shortly after the U.S. government temporarily restricted access to Anthropic’s Fable 5 and Mythos 5 models following a jailbreak that exposed access to offensive cybersecurity capabilities. Although those restrictions were lifted on July 1, the incident highlighted how quickly AI security concerns are becoming a matter of national importance.
The scale of cyberattacks has changed, and the stakes have never been higher. Identity represents the single greatest point of leverage. Knowing who and what is accessing your systems, continuously and verifiably, is the main factor in preventing an AI-powered attack or potentially leading to a serious breach.
As the Five Eyes wrote: “Cyber risk can no longer be treated as a purely technical issue. This is a core business risk and leadership responsibility.” I’ve been singing the same tune for a long time. This isn’t about compliance or crossing off the items on a basic checklist. It’s time for governments and businesses the world over to recognize what this caliber of cyber risk represents: we must figure out how to manage identity and its far-reaching effects.
It’s also time to get serious about non-human identities. They now vastly outnumber human identities, and the rapid rise of agentic AI is transforming them from passive, deterministic processes into autonomous digital actors capable of making decisions and initiating actions at machine speed.
Every AI agent that operates in your environment requires a governance framework that can verify who authorized it, what it’s permitted to do, and whether it’s still operating within that scope. Such agents should also have bound tokens that can be audited and traced back to a human. We also need to establish lines of accountability as an industry. Who is accountable when an AI agent acts on your behalf? And how do you govern an identity that can replicate, reason, and act independently, often without human oversight?
The greatest concern is that our industry conversations surrounding agentic identity governance and verifiable credential ecosystems have very little to do with what’s actually being deployed in the outside world.
These criminals didn't break through firewalls. Instead, they walked right through the proverbial front door via hiring processes that relied on resume screening, video calls, and other forms of verification that can be easily defeated. They also used generative AI to forge thousands of synthetic identities, alter photos, mask their appearances during video interviews, and answer technical coding questions in real time.
However, bad actors live everywhere; this is about a lot more than North Korea. And the window to build the right foundation is narrowing fast. The world still largely runs on passwords, SMS codes, and so-called secret questions about concerts and maiden names. Not only are these not secure, but they also were not designed for AI-powered threats. The scary part is that in some cases we are legally mandating them, even though our own standards bodies have deemed them insecure. As we usher in the near future, this reality should be regarded for what it is: a five-alarm crisis.
The concept is simple: Circle of Identity assures a continuous relationship between a person and the institutions, platforms, and systems that need to verify who they are, across many interactions, over the course of their relationship. This distinction matters because most attacks happen in the gaps between verification events.
Today, those gaps are everywhere. A customer may be verified when opening an account, but when they replace a device, call a service center, or recover their credentials, organizations often fall back on passwords, knowledge-based questions, or information that is already available on the dark web. The original verification and subsequent interactions are rarely connected, creating opportunities for fraudsters to exploit.
A closed Circle of Identity operates very differently. When a foundational biometric-anchored identity is established at enrollment, that verification becomes the persistent reference point for every subsequent interaction. Device provisioning, account recovery, step-up authentication, and high-risk transactions all trace back to that original verification, preserving continuity and dramatically reducing opportunities for account takeover and impersonation.
This foundation is particularly important as organizations embrace agentic AI and digital credentials. These technologies represent the future of digital trust, but they depend on strong identity assurance at the human level. You can’t build a reliable credential ecosystem if individuals can obtain multiple credentials under different identities. You can’t govern AI agents without confidently verifying the humans who authorize and oversee them.
The Five Eyes alliance warned organizations to act now and be prepared for AI-enabled cyber threats. I’ve been saying the same thing with less authority, but the same urgency, for a long time. As technology continues to evolve, the principle remains unchanged: trust begins with knowing, continuously and verifiably, who is on the other side of every interaction.
Will we transform our identity management strategies before it’s too late, or are we willing to risk it all?
About author: Frances Zelazny is the General Manager of New Market Initiatives at Prove. She leads the development and commercialization of Prove’s new privacy-preserving biometric and KYC compliance solutions.
Reviewed by Irfan Ahmad.
Read next:
• AI can be a personal trainer in your pocket – but is it safe?
• Many Teenagers Show Symptoms of Excessive Screen Use
The Five Eyes intelligence alliance between the U.S., U.K., Canada, Australia, and New Zealand recently issued a rare joint statement: the potential for devastating, AI-powered cyberattacks is months (not years) away. The Five Eyes statement comes shortly after the U.S. government temporarily restricted access to Anthropic’s Fable 5 and Mythos 5 models following a jailbreak that exposed access to offensive cybersecurity capabilities. Although those restrictions were lifted on July 1, the incident highlighted how quickly AI security concerns are becoming a matter of national importance.
The scale of cyberattacks has changed, and the stakes have never been higher. Identity represents the single greatest point of leverage. Knowing who and what is accessing your systems, continuously and verifiably, is the main factor in preventing an AI-powered attack or potentially leading to a serious breach.
AI-Powered Cyber Threats Create Outsized Concerns
The timing between the Anthropic news and the Five Eyes statement are no coincidence. Let’s examine the Anthropic issue a little further. Though the U.S. government initially ordered access control based on nationality, that approach wasn’t something Anthropic could achieve because there is no way to ascertain that for most Americans who don’t hold a U.S. passport, let alone others from around the world. Since the AI company couldn’t enforce restrictions selectively, access to Fable 5 and Mythos 5 was temporarily suspended before being restored on July 1.As the Five Eyes wrote: “Cyber risk can no longer be treated as a purely technical issue. This is a core business risk and leadership responsibility.” I’ve been singing the same tune for a long time. This isn’t about compliance or crossing off the items on a basic checklist. It’s time for governments and businesses the world over to recognize what this caliber of cyber risk represents: we must figure out how to manage identity and its far-reaching effects.
The Time to Shore Up Security is Now
Despite risk, the vast majority of governments and organizations have continued to take their chances on cybersecurity methods that no longer fit the bill. Bad actors’ methods evolve, as should our approaches to identity management. First, it’s time to eliminate phishable credentials from your authentication stack: passwords, OTPs, and push notifications are now AI-friendly attack surfaces. Second, it’s time for a layered approach, for example, privacy-preserving biometrics bound to trusted devices augmented by intelligence and dynamic signals for ongoing, verifiable identification at scale.It’s also time to get serious about non-human identities. They now vastly outnumber human identities, and the rapid rise of agentic AI is transforming them from passive, deterministic processes into autonomous digital actors capable of making decisions and initiating actions at machine speed.
Every AI agent that operates in your environment requires a governance framework that can verify who authorized it, what it’s permitted to do, and whether it’s still operating within that scope. Such agents should also have bound tokens that can be audited and traced back to a human. We also need to establish lines of accountability as an industry. Who is accountable when an AI agent acts on your behalf? And how do you govern an identity that can replicate, reason, and act independently, often without human oversight?
The greatest concern is that our industry conversations surrounding agentic identity governance and verifiable credential ecosystems have very little to do with what’s actually being deployed in the outside world.
Sometimes the Threat Is Already in the Building
If you can’t continuously verify the identity of the humans and machines that touch your infrastructure, the rest of your efforts are the equivalent of securing the perimeter against an adversary who’s already entered the building. Christina Chapman, an American woman, was sentenced to more than eight years in federal prison for helping North Korean IT workers gain employment at more than 300 U.S. organizations, including government agencies, using the stolen identities of 68 Americans. The Justice Department called it the largest identity-theft case of its kind. Since then, the problem has only escalated. CrowdStrike's 2025 Threat Hunting Report identified more than 320 incidents over the past 12 months, a 220% year-over-year rise, through Famous Chollina alone, in which North Koreans gained fraudulent employment at Western companies working remotely as developers.These criminals didn't break through firewalls. Instead, they walked right through the proverbial front door via hiring processes that relied on resume screening, video calls, and other forms of verification that can be easily defeated. They also used generative AI to forge thousands of synthetic identities, alter photos, mask their appearances during video interviews, and answer technical coding questions in real time.
However, bad actors live everywhere; this is about a lot more than North Korea. And the window to build the right foundation is narrowing fast. The world still largely runs on passwords, SMS codes, and so-called secret questions about concerts and maiden names. Not only are these not secure, but they also were not designed for AI-powered threats. The scary part is that in some cases we are legally mandating them, even though our own standards bodies have deemed them insecure. As we usher in the near future, this reality should be regarded for what it is: a five-alarm crisis.
The Circle of Identity Way, Continuous and Verifiable
With more than 30 years of witnessing urgency, breakthroughs, and brilliant standards accompanied by complacency, slow adoption, and partial implementation, I keep coming back to the same fundamental truth: we must maintain persistent identity across the user lifecycle in every service channel. This means threading humans through enrollment, device registration, authentication, and account recovery, whether on the phone, in person, online, or via a chat or agent. I call this the Circle of Identity.The concept is simple: Circle of Identity assures a continuous relationship between a person and the institutions, platforms, and systems that need to verify who they are, across many interactions, over the course of their relationship. This distinction matters because most attacks happen in the gaps between verification events.
Today, those gaps are everywhere. A customer may be verified when opening an account, but when they replace a device, call a service center, or recover their credentials, organizations often fall back on passwords, knowledge-based questions, or information that is already available on the dark web. The original verification and subsequent interactions are rarely connected, creating opportunities for fraudsters to exploit.
A closed Circle of Identity operates very differently. When a foundational biometric-anchored identity is established at enrollment, that verification becomes the persistent reference point for every subsequent interaction. Device provisioning, account recovery, step-up authentication, and high-risk transactions all trace back to that original verification, preserving continuity and dramatically reducing opportunities for account takeover and impersonation.
This foundation is particularly important as organizations embrace agentic AI and digital credentials. These technologies represent the future of digital trust, but they depend on strong identity assurance at the human level. You can’t build a reliable credential ecosystem if individuals can obtain multiple credentials under different identities. You can’t govern AI agents without confidently verifying the humans who authorize and oversee them.
The Five Eyes alliance warned organizations to act now and be prepared for AI-enabled cyber threats. I’ve been saying the same thing with less authority, but the same urgency, for a long time. As technology continues to evolve, the principle remains unchanged: trust begins with knowing, continuously and verifiably, who is on the other side of every interaction.
Will we transform our identity management strategies before it’s too late, or are we willing to risk it all?
About author: Frances Zelazny is the General Manager of New Market Initiatives at Prove. She leads the development and commercialization of Prove’s new privacy-preserving biometric and KYC compliance solutions.
Reviewed by Irfan Ahmad.
Read next:
• AI can be a personal trainer in your pocket – but is it safe?
• Many Teenagers Show Symptoms of Excessive Screen Use
