These are interesting times for AI and trust. A growing number of investment firms are using AI agents to review research notes and corporate documents. People are being asked to hand over increasingly invasive biometric data, such as facial scans, voice samples and behavioral patterns, just to prove they aren’t bots. Once in the wild, this data can be weaponized by AI-driven bots to convincingly spoof real people, defeating the systems designed to keep them out. That puts us in a strange new arms race: the more invasive the verification, the greater the risk when it inevitably leaks. So, how do we verify who (or what) we are really dealing with?
It is unconscionable to demand transparency from people while accepting opacity from machines. Both bots and online humans need better ways to verify their identities. We can’t solve this problem by simply collecting more biometric data, nor by setting up centralized registries that create massive honeypots for cybercriminals. Zero-knowledge proofs offer a path forward where both humans and AI can prove their credentials without exposing themselves to exploitation.
The trust deficit blocks progress
The lack of a verifiable AI identity creates immediate market risks. When AI agents can impersonate humans, manipulate markets or carry out unauthorized transactions, companies are rightly hesitant to deploy autonomous systems on a large scale. In fact, LLMs that have been ‘refined’ on a smaller data set to improve performance are 22 times more likely to produce harmful results than base models, with success rates of bypassing the system’s security and ethical guardrails – a process known as ‘jailbreaking’ – tripling compared to production-ready systems. Without reliable identity verification, every AI interaction brings you one step closer to a potential security breach.
The problem isn’t as obvious as preventing malicious actors from deploying rogue agents, because it’s not like we’re dealing with a single AI interface. In the future, there will be more and more autonomous AI agents with greater capabilities. In such a sea of agents, how do we know what we are dealing with? Even legitimate AI systems need verifiable credentials to participate in the emerging agent-to-agent economy. When an AI trading bot executes a trade with another bot, both parties need certainty about the other’s identity, authorization, and accountability structure.
The human side of this equation is also broken. Traditional identity verification systems expose users to massive data breaches, allow too-easy authoritarian surveillance, and generate billions in revenue for major corporations from selling personal information without compensating the individuals who generate it. People are rightly reluctant to share more personal data, but legal requirements require increasingly intrusive verification procedures.
Zero-Knowledge: the bridge between privacy and accountability
Zero-knowledge proofs (ZKPs) offer a solution to this seemingly intractable problem. Rather than revealing sensitive information, ZKPs allow entities, both human and artificial, to prove specific claims without exposing underlying data. A user can prove that he is over 21 without revealing his date of birth. An AI agent can prove that it has been trained on ethical data sets without exposing proprietary algorithms. A financial institution can verify that a customer meets regulatory requirements without storing personal information that could be compromised.
For AI agents, ZKPs can enable the necessary deep levels of trust, as we need to verify not only the technical architecture, but also behavioral patterns, legal accountability, and social reputation. ZKPs allow these claims to be stored in a verifiable on-chain trust graph.
Think of it as a composable identity layer that works across platforms and jurisdictions. That way, when an AI agent presents its credentials, it can prove that its training data meets ethical standards, that its results have been audited, and that its actions are linked to responsible human entities, all without exposing proprietary information.
ZKPs could completely change the game, allowing us to prove who we are without transferring sensitive data, but adoption remains slow. ZKPs remain a technical niche, unknown to users, and mired in regulatory gray areas. Furthermore, companies that benefit from data collection have little incentive to adopt the technology. However, that hasn’t stopped more flexible identity companies from taking advantage. As regulatory standards emerge and awareness improves, ZKPs can become the backbone of a new era of trusted AI and digital identity – giving individuals and organizations a way to communicate with each other securely and transparently across platforms and borders.
Implications for the market: unlocking the agent economy
Generative AI could add trillions annually to the global economy, but much of this value remains hidden behind identity verification barriers. There are several reasons for this. One is that institutional investors need robust KYC/AML compliance before deploying capital into AI-driven strategies. Another is that companies need verifiable agent identities before autonomous systems can access critical infrastructure. And regulators are demanding accountability mechanisms before approving the use of AI in sensitive domains.
ZKP-based identity systems meet all these requirements while maintaining the privacy and autonomy that make decentralized systems valuable. By enabling selective disclosure, they meet legal requirements without creating a honeypot of personal data. By providing cryptographic authentication, they enable reliable interactions between autonomous agents. And by maintaining user control, they align with new data protection regulations such as GDPR and California privacy laws.
The technology could also help tackle the growing deepfake crisis. When every piece of content can be cryptographically linked to a verified creator without revealing their identity, we can fight disinformation and protect privacy. This is especially critical as AI-generated content becomes indistinguishable from human-generated material.
The ZK path
Some will argue that any identity system represents a step toward authoritarianism — but no society can function without a way to identify its citizens. Identity verification is already happening on a large scale, just on a limited scale. Every time we upload documents for KYC, undergo facial recognition, or share personal data for age verification, we are participating in identity systems that are invasive, insecure, and inefficient.
Zero-knowledge proofs offer a path forward that respects individual privacy while enabling the trust necessary for complex economic interactions. They allow us to build systems where users control their data, authentication requires no supervision, and both humans and AI agents can communicate securely without sacrificing autonomy.