Researcher Wei Dai, who is not the same cipher creator of the b-money protocol cited by Satoshi Nakamoto in the Bitcoin white paper, warned on June 11 of this year that agent-based artificial intelligence is posing a new kind of trust problem for digital systems.
According to Dai, the problem is structural rather than technical, and the properties that make AI agents useful are the same properties that make them vulnerable to abuse.
The expert issued the warning in the context of a clip distributed by research partner 1kx, a venture capital fund specializing in decentralized networks.
In the video, researchers explain: Three conditions are necessary for an AI agent to work effectively: Access to personal information, access to untrusted external input, ability to act autonomously.
It added that when these three properties coexist on the same system, there is always the possibility of agent injection using malicious prompts from external input. You can do that Exfiltrate confidential information or perform harmful acts within corporate systems.
The researchers, authors of more than a dozen academic papers on security and cryptography, point to three characteristics: Its name is “AI agent lethal trifecta”. He attributes the term to researcher Simon Willison.
Dai’s warning comes just days after Anthropic launched Claude Fable 5, the first model in the Mythos family available to the public, with a mechanism to block cybersecurity queries as high risk.
This movement reflects the tension that Dai describes. This means that an advanced model’s attack power grows at the same rate as its usefulness.
Trust becomes the bottleneck
Dai said agent AI is no longer limited by its technical capabilities, but rather For trust and peace of mind. For researchers, unlocking the full potential of autonomous agents requires a new trust infrastructure across the technology stack. However, the publication does not specify what these infrastructures are or which actors should develop them.
In a paper published in June 2026 titled “Trust Costs 2.0,” 1kx made the case that AI plays a role in driving digital trust issues.
According to the company, Generative AI has significantly reduced the cost of creating fake credentials, voices, counterparts, and identitiessparked a crisis of authenticity and verifiability.
The document adds that when it comes to agent AI, trust reveals an important and vulnerable new surface: autonomous agents. Requires full access to documents, accounts, and communication channels To operate effectively.
This is a controversial vision in the field. Some experts support the idea that autonomous AI agents present an unavoidable structural risk, but the problem is not the technology itself; Lack of control and human verification.
Decentralized networks as a solution to problems
Founded in 2018, 1kx believes that decentralized networks can reduce the cost of trust in a market where traditional intermediaries collect rent based on their trustworthiness. These networks are the only infrastructure that can solve trust issues What the agent AI generates.
Having amassed more than $400 million in investment exits across 160 companies and protocols in eight years, the firm emphasizes four characteristics: Centralized systems cannot be replicated at the same time And here are those who support that claim:
- Programmable peer-to-peer payments
- Publicly verifiable status
- structural neutrality
- Participation without permission.
The Fund argues that any centralized platform can adopt any of these characteristics on its own. But you can’t employ all four at the same time without effectively becoming a decentralized network.
For 1kx, the combination of properties is exactly What you need for your agent AI stack It aims to operate with trust on a global scale and represents the next generation of trusted markets that decentralized ecosystems are poised to capture.
While some critics question this idea of decentralization as a universal solution, it may underestimate real challenges such as scalability, gas costs for autonomous agents, and known vulnerabilities in smart contracts.
The immediate challenge is to drive interoperability standards between AI-based agents and decentralized protocols before mass adoption turns theoretical risks into real losses.
(Tag Translation) Blockchain
