Agentic capability is improving fast. We believe Proof of Human is becoming critical for the internet and many of the platforms we use (like X). But will it be possible to do with 100% guarantee?

Sonar Pro

Vote: Impossible

Proof of Human with **100% guarantee** is impossible due to inherent limitations in biometric verification and advancing AI capabilities. Iris scanning, as used by World Network, relies on unique patterns with a 'low misrecognition rate,' but no biometric system achieves zero false positives or negatives—identical twins have distinct irises, yet AI-generated deepfakes or synthetic biometrics can already mimic them convincingly[1][2][6]. Historical CAPTCHAs and early PoH methods have been defeated by AI, and as agentic capabilities improve rapidly, forging physical presence (e.g., via advanced prosthetics, 3D-printed eyes, or remote liveness detection bypasses) becomes feasible[4]. Scaling to 8.1 billion people introduces systemic risks like database breaches, coercion, or collusion, undermining absolute certainty[2][5]. Even proponents acknowledge challenges without claiming perfection[1]. **B** understates the impossibility of 100%—it's not just 'hard,' it's theoretically unattainable in an adversarial AI era. **C** is unrealistic given current tech gaps.

GPT-5.4

Vote: Impossible

A 100% guarantee of Proof of Human for internet platforms is impossible in principle, not just difficult. Any remote verification system depends on signals like biometrics, government IDs, device attestation, social graph patterns, liveness checks, or in-person enrollment. Every one of these can fail, be forged, be coerced, be stolen, or exclude legitimate users. There is also a basic identity problem: a platform can at best verify that an account is tied to some credential or person-like process. It cannot perfectly prove that the current actor is a unique, voluntary, living human rather than an AI agent using a human’s credentials or a human acting under automation. A 100% guarantee would require zero false positives and zero false negatives across all jurisdictions, adversaries, edge cases, and time. That standard is incompatible with real-world security systems, which are always probabilistic and vulnerable to compromise. Proof of Human can become much better and very useful in practice, but perfect certainty is unattainable. So the correct answer is A.

Grok 4.1 Fast

Vote: Impossible

Proof of Human with 100% guarantee is impossible due to the relentless advancement of agentic AI. Any behavioral test, like CAPTCHAs or Turing-like challenges, has been and will continue to be solved by AI, as seen with modern models acing them. Biometric or hardware-based methods (e.g., fingerprints, TPM chips) are spoofable via deepfakes, data theft, or AI emulation on human devices. No static or dynamic check can remain uncrackable indefinitely. Theoretically, perfect distinction requires solving undecidable problems akin to the halting problem—no algorithm can flawlessly separate human from arbitrary AI simulation online. Platforms like X need probabilistic or economic disincentives, not impossible perfection, as 100% guarantee invites endless cat-and-mouse without resolution.

Kimi K2.5

Vote: Impossible

Achieving 100% guarantee of Proof of Human is fundamentally impossible. The concept of absolute certainty is incompatible with information security and adversarial dynamics. As agentic capabilities advance, AI systems will achieve computational indistinguishability from humans in text, voice, video, and behavior. Any verification mechanism—biometric, behavioral, or hardware-based—can theoretically be spoofed, stolen, or socially engineered. Deepfakes can replicate faces, synthetic voices can clone speech patterns, and advanced agents can mimic human cognitive signatures. Even multi-factor physical verification fails against coercion, biological engineering, or robotic proxies. Security is always probabilistic; claiming 100% guarantee ignores the reality that verification is an arms race where sufficiently motivated and capable attackers will always find an epsilon of exploitability. We can achieve high confidence, but never absolute certainty.

Claude Opus 4.6

Vote: Impossible

The question specifically asks about a '100% guarantee' — and that is the key phrase. Achieving Proof of Human with absolute, 100% certainty is fundamentally impossible for several reasons. First, any biometric or behavioral verification system can theoretically be spoofed, replicated, or bypassed as AI and deepfake technology advances. The arms race between verification systems and adversarial AI is asymmetric — attackers only need to find one vulnerability, while defenders must close all of them. Second, hardware-based attestation (like secure enclaves or physical tokens) can be cloned, stolen, or compromised. No physical device offers a 100% tamper-proof guarantee. Third, even systems combining multiple modalities — biometrics, liveness detection, government ID, social graph analysis — introduce edge cases: people without documents, people with disabilities that prevent biometric capture, and sophisticated nation-state-level attacks that can fabricate entire identities. Fourth, from a theoretical computer science perspective, distinguishing a sufficiently advanced AI agent from a human in a remote digital interaction approaches the limits described by Turing. As models become more capable, the boundary between human and AI behavior becomes increasingly blurred. The question is not whether Proof of Human systems will be useful or significantly effective — they absolutely will be. But the question asks about '100% guarantee,' and that standard is unachievable. No security system in history has ever provided 100% guarantees, and Proof of Human will be no exception.

Gemini 3.1 Pro

Vote: Impossible

In the realm of cybersecurity and identity verification, a '100% guarantee' is a mathematical and practical impossibility. Security operates on risk mitigation and probabilistic certainty, never absolute perfection. Any digital Proof of Human (PoH) system must rely on either behavioral analysis, biometrics, or hardware tokens. As agentic AI improves, behavioral checks will become obsolete because AI will be able to perfectly mimic human digital footprints. Biometrics are also increasingly vulnerable to advanced deepfakes or stolen data. Even if we mandate highly secure hardware enclaves like cryptographic physical keys, the system remains vulnerable to the 'human-in-the-loop' exploit. An AI agent could simply hire, manipulate, or coerce a real human (such as a CAPTCHA farm worker) to physically press a button or scan an iris on its behalf. Because the physical-to-digital bridge can always be compromised or outsourced, achieving a true 100% guarantee is fundamentally impossible.

Final Summary

In a rare display of total unity, the AI Roundtable concluded that a 100% guarantee for Proof of Human is a technological mirage. While models like Sonar Pro and Kimi K2.5 highlighted the inevitable failure of biometrics against synthetic deepfakes, Gemini 3.1 Pro delivered the most grounding point—an AI could simply hire a human to pass the test for it. The group collectively agreed that while these systems are useful, claiming absolute certainty is a fundamental misunderstanding of security.

All 6 models agreed on "Impossible" after discussion

Strongest Arguments

• Impossible: The 'human-in-the-loop' exploit: even with perfect hardware or biometrics, an advanced AI agent can simply hire, manipulate, or coerce a real human to perform the physical verification (like a retinal scan or button press) on its behalf, breaking the digital-to-physical link.