World ID can prove a real human is behind an account. It can't prove that human hasn't already been phished, vished, or bribed. The biggest breaches of the past three years didn't fake humanity, they exploited it.
In September 2023, Scattered Spider attacked MGM Resorts. They didn’t deploy zero-day exploits. They didn’t crack encryption. They called the IT helpdesk, spent ten minutes on the phone impersonating an employee they’d found on LinkedIn, and walked straight into Okta and Azure.
$100 million in losses. 37 million customer records exposed. A $45 million class-action settlement. Zero technical exploits.
The attacker talked to a real, verified, authenticated human being the entire time. That was the point.
A new category of identity technology is going mainstream in 2026. Proof of Human platforms, led by World ID (built by Tools for Humanity, co-founded by Sam Altman), promise to solve one of the internet’s most pressing problems: the inability to distinguish real people from bots, AI agents, and synthetic identities.
It’s a real problem. GPT-4 passed the Turing Test. Agentic AI systems now conduct full attack chains autonomously. Deepfakes pass KYC checks. CAPTCHAs, which were never designed for a world where AI outperforms humans at puzzle-solving, are effectively dead as a trust mechanism.
World ID’s response: scan your iris at a physical “Orb” device. Generate a cryptographic proof of uniqueness stored on your phone, anchored to Ethereum’s World Chain. Present that credential to any app or service, via zero-knowledge proof, without revealing personal data. As of March 2026, 18 million verified humans have done exactly this across 160+ countries. World ID is live on Tinder, Shopify, and Discord.
The cryptography is serious. The privacy architecture, at least in theory, is careful. And the problem it’s solving, bot proliferation, AI impersonation, Sybil attacks, is genuinely worth solving.
But there’s something these platforms don’t claim to do. And it’s the thing that matters most to security leaders.
Proof of Human tells you a real, unique person is behind an account. It tells you nothing about whether that person has been phished, vished, bribed, coerced, or is in the process of being socially engineered right now.
Verizon’s 2025 Data Breach Investigations Report is unambiguous: 60% of all data breaches in 2025 involved the human element. Not spoofed identities. Not AI impersonation. Real people, making decisions under manipulation, pressure, or deception.
The numbers from 2025 make this hard to ignore:
The attacker doesn’t need to fake humanity. They’ve moved on. They just need to find the real human and work on them.
This isn’t theoretical. The biggest breaches of the past three years were not won by AI impersonation or bot infiltration. They were won through social engineering of real, verified, authenticated people. This is the human factor in cybersecurity, and no identity platform addresses it.
Scattered Spider researched an MGM employee on LinkedIn, called the IT helpdesk, and persuaded a real employee to reset credentials and hand over access to Okta and Azure. Ten minutes. $100 million in losses. vx-underground put it plainly at the time: “A company valued at $33,900,000,000 was defeated by a 10-minute conversation.”
Lapsus$ bought contractor credentials on the dark web. The contractor had MFA. Lapsus$ bombarded them with push notifications, then WhatsApp-messaged the contractor directly, posing as Uber IT security, telling them to approve the flood. The contractor, a real, employed, verified human, complied. Full access to AWS, Slack, GCP, SentinelOne, and the company’s bug bounty platform admin followed.
Scattered Spider targeted TCS, M&S’s outsourced IT helpdesk provider. They impersonated employees and social-engineered TCS staff into resetting passwords and granting system access, bypassing MFA. They spent weeks in the network before deploying ransomware over Easter weekend. M&S lost approximately £300 million in profits (M&S Annual Results, May 2025). Market cap dropped by over £1.2 billion. Ecommerce was down for 46 days. Four individuals aged 17 to 20 were later arrested in the UK. The River Island Information Security Officer, speaking at Infosecurity Europe 2025, described the techniques as “elegant and subtle, but not as complicated as we imagine.”
Threat actors didn’t impersonate anyone to a system. They bribed verified, employed contractors to use their legitimate access to export customer records. No technical exploit. No fake identity. Just money, paid to real humans with real access. Coinbase CSO Philip Martin described it as “an evolution in attacker behavior”: “I’ve never heard of the kind of pervasive bribery that this incident showed us, with the long-term focus and the amounts involved.” Estimated remediation cost: $180 million, per Coinbase’s own public disclosure.
Each of these involved verified humans. Employed, authenticated, real people. The verification wasn’t the problem. It was the precondition for the attack.
If anything, Proof of Human platforms create a new dependency: once you’ve proven you’re human, systems extend trust. The credential becomes a key. And keys can be stolen, sold, rented, or coerced out of their owners.
This isn’t speculation. Vitalik Buterin raised it directly in a July 2023 analysis of biometric proof of personhood: “If it’s common knowledge that everyone has only one identity, you can be coerced into revealing it.” He identified credential selling and renting as structural weaknesses that zero-knowledge proofs cannot fix. Glen Weyl, economist at Microsoft Research, was more direct: “You’ve just found a way to generate a key, and that key can be sold or disposed of in any way people want.”
Research published in the Stanford Journal of Blockchain Law and Policy in January 2025 documented this in practice. On Idena, a competing proof-of-personhood protocol, researchers found evidence of “puppeteering,” pool operators controlling verified accounts from the moment of registration. The same paper documented a black market for World IDs operating out of China, Cambodia, and Kenya, with verified credentials bought and sold as commodities.
Even Worldcoin’s own whitepaper acknowledged: “It is expected that the Orb may get spoofed or compromised by determined actors.”
The platform designed to verify humanity is not designed to protect it. That’s not a criticism of the technology. It’s a precise statement of scope.
The same logic applies to MFA, the technology most organizations already rely on as their primary identity assurance layer.
The 2025 data tells a consistent story:
Only 33% of security leaders say they’re confident their identity provider can prevent identity-based attacks (Cisco 2025). That’s despite identity-based attacks accounting for 60% of all Cisco Talos incident response cases in 2024. Phishing-resistant authentication methods like hardware security keys raise the bar. They don’t change the calculus when the human holding the key is the target.
Ontinue’s 2H 2025 Threat Intelligence Report captured it clearly: “Attackers increasingly behave like sophisticated users rather than system intruders… Once authenticated, even improperly, they inherit the trust the organization has given to that identity. Identity is the new perimeter because it’s the only boundary attackers must cross to behave like insiders.”
Verification is a moment. Attackers operate across time.
The gap between identity verification and security posture is not a technology problem that a better verification system will solve. It’s a behavioral and operational problem. And the only way to find it before an attacker does is to simulate the attacker.
Not annually. Not on a scoped engagement. Continuously, across every surface: digital, physical, and social.
A vishing simulation tests whether your IT helpdesk gives up credentials to a convincing caller. A social engineering prevention program requires you to first understand exactly how and where your people can be manipulated. Spear phishing and vishing simulations surface that before an attacker does. A red team exercise tests whether an attacker, starting with nothing but LinkedIn and a phone, can reach your most sensitive systems the way Scattered Spider did at MGM, M&S, and Coinbase.
Proof of Human platforms are solving the bot problem. That’s legitimate work. But the threat has already adapted. Attackers aren’t trying to fake humanity anymore. They’re investing in finding the humans who can be worked, bribed, worn down, or deceived into handing over access.
Lapsus$ documented their methodology publicly on Telegram:
“No limit is placed on the number of calls that can be made. Call the employee 100 times at 1am while he is trying to sleep, and he will more than likely accept it.”
They weren’t wrong.
If you’re thinking about Proof of Human platforms, the right question isn’t “does this verify identity?” It’s what comes after.
These aren’t questions World ID can answer. They’re questions a red team should be answering, continuously, before an attacker does it for real.
The threat of cyber attack is constant. So are we. Schedule a call to discuss how to outpace cyber threats.
Industry Reports
Academic & Expert Commentary
Company Disclosures
Proof of Human solves the bot problem. It doesn’t solve the people problem.
World ID can prove a real human is behind an account. It can’t prove that human hasn’t already been phished, vished, or bribed. The biggest breaches…
Too many rules, no real test: Untangling US Cyber Disclosure
The US has no single federal data breach notification law, just a growing patchwork of SEC rules, HIPAA, state obligations, and incoming CIRCIA requirements. Most organizations…
Project Glasswing is impressive. But what about the rest?
Anthropic’s Project Glasswing is a serious step forward for technical security. But it covers one third of the attack surface. Here’s what it leaves out, and…