Empowered by the interoperability introduced by blockchain smart contracts, users can perform various operations on-chain. After connecting your wallet, you can transfer tokens, swap tokens, bridge tokens, and engage in lending and redemption. You can even create your own smart contracts for specific purposes. These features make operations on the blockchain more flexible and diverse.
According to Trusta's analysis of on-chain Sybils, at least 20% of on-chain wallets are Sybils, contributing over 40% of on-chain activities. The remaining activities are generated by humans. Thus, crypto activities are primarily driven by (1) Human Intelligence and (2) repetitive bots (Sybils).
Despite the significant attention given to emerging and high-valued AI agents in the crypto world, cases of on-chain operations based on intelligent autonomous AI agents are still in the early stages. Therefore, the contribution of on-chain activities by Artificial Intelligence is currently negligible. However, this landscape will soon be disrupted (within 1-2 years). Trusta believes that before 1 billion humans onboard to crypto, we will see a scenario where 1 billion AI agents achieve mass adoption on-chain.
Before we provide the reasons, let's first look at the definitions and distinctions of these three types of entities.
The table compares the three entities: Human intelligence, programmed bots, and AI agents.。
Description
Decision-Making
Action-Taken
Intelligence Level
Human
This specifically refers to human participation in blockchain activities in person. Actions such as connecting wallets, swapping, bridging, lending, and trading alpha are all initiated by humans.
All decisions are made by humans, and the entire workflow is planned based on human decisions and intelligence.
hands-on actions by humans
Human-level intelligence
Bots
AKA Sybils. One human controls and manipulates a bunch of bots to interact with the blockchain. The main purpose of Sybils is to farm airdrops and rewards from projects.
All decisions and actions are determined by pre-specified logic configured by humans.
Script for a computer program
Zero intelligence
AI Agents
In an ideal scenario, with user approval, they can independently make decisions and engage in on-chain interactions on their behalf, guided by specific intentions.
Intent-centric； Decisions are made based on the context and user identity through the LLM.
Tools for AI agent
The backed AI models are on par with average human levels.
The activity level of on-chain AI agents is expected to see explosive growth in the next 2-3 years (by 2028). Autonomous Artificial Intelligences will completely replace simple repetitive Sybil behavior and even some Human Intelligence (as everyone will have the ability to create their own AI agents). It is anticipated that they will dominate at least 80% of on-chain activities, becoming the primary force in on-chain activities, while Humans and Sybils will account for the remaining 20%.
AI + Crypto has become a buzzword and a hot investment trend. It is believed that AI agents will ultimately dominate the crypto world for FOUR reasons.
The infrastructure for AI+Crypto is rapidly evolving. The advancements in AI agent frameworks, crypto tools for AI agents, multi-agent orchestration frameworks, and other supporting facilities, such as AI agent payments, are becoming increasingly sophisticated. These enhancements not only improve the capabilities of AI agents but also significantly lower the barriers to building them. As a result, each user is expected to have at least five on-chain AI agents to serve their needs. Moreover, AI agents could even hire other agents, including humans, to fulfill more complex tasks.
AI agents align perfectly with the cryptocurrency world's demand for intelligence. They are simple, trustworthy, and secure, making them ideal for this environment. In the crypto realm, where "code is law" and trustlessness are prioritized, there is a strong preference for communication, interaction, and transactions driven by AI models and code. In this context, AI agents resemble Trisolarans, possessing great intelligence without the tendency to deceive others. As a result, the demand for AI agents in the cryptocurrency world is expected to grow even stronger.
AI agents will revolutionize human interaction with blockchain. They will shift the process from the cumbersome and insecure "Connect Wallet" method to an intent-centric approach of approving AI agents. The intelligence of advanced beings is expressed in an intent-centric manner; for example, a human's interaction with DeFi starts with the intention, "I need a relatively safe DeFi protocol to achieve an APY of over 4% with my assets in USDC," rather than the mechanical steps of connecting a wallet, swapping tokens, approving tokens, and providing liquidity. AI agents will facilitate intent-centric on-chain interactions for humans, significantly reducing friction and aiding in mass adoption.
AI-driven applications excel at capturing public attention and traffic. Additionally, AI agent tokenization gives them a natural advantage in capturing value.
So, let’s assume that we have an ecosystem dominated by AI agents, coexisting with human intelligence, and also featuring a small number of Sybils. What kinds of challenges related to trusted identity for crypto intelligence will need to be addressed? Please note that the challenges listed below are only those currently identifiable and needing resolution. As Trusta continues to explore and tackle these issues, it is likely that new problems will emerge.
World ID addresses the challenges of Universal Basic Income (UBI) by implementing a Proof of Human (PoH) system. This ensures that only real individuals can access UBI benefits, preventing fraud and misuse. Trusta.AI also has substantial work focused on human verification to distinguish between humans and Sybils based on AI algorithms. Sybils merely replicate the owner's instructions in a simple and deterministic manner, lacking any intelligence, let alone autonomy. Excessive Sybil behavior only provides a numerical boost and offers no real benefits for project development or ecosystem growth.
However, PoH merely distinguishes between humans and Sybils, which is far from sufficient. In the era of AI, AI agents are fundamentally different from Sybils, as they are autonomous, intelligent entities, meaning they possess autonomy and the ability to make self-directed decisions. Therefore, Trusta believes that:
In the era of AI, all intelligences (humans and AI agents) deserve rewards.
Thus, the challenge in the AI era is to accurately distinguish Sybils and grant legitimate identities to both humans and AI agents.
Trusta.AI is focusing on decentralizing its identity service to foster a self-sovereign identity ecosystem. This will support multiple identity verification methods, including document verification, biometrics, and AI-driven approaches.
To meet scalability demands, Trusta.AI is developing an open, trustless, and verifiable Identity layer, known as the Trusta Attestation Service (TAS). Key features of TAS include:
Diverse Verification Methods: TAS supports various identity verification methods, allowing attestations through AI-driven techniques, documents, biometrics, and more.
Single Attestation: Users will only need to attest once, enabling them to use their verified identity across different platforms and services.
This approach is designed to enhance user convenience while ensuring robust identity verification.
The Sybils score and the reputation assessment are two orthogonal dimensions to evaluate an account. One is used to select authentic users, while the other is used to rank high-value users. In the Crypto+AI world, we need an objective, fair, and quantifiable metric to comprehensively evaluate human and AI agents' on-chain engagement and value. The MEDIA Score has been developed to enable projects to accurately target users who have truly contributed to the project and ensures that resources and incentives are fairly distributed to these users. In 【10】, Trusta has published its work on account selection of zkSync organic users based on the intersection of two dimensions. A similar methodology can be used to assess the value of AI agents, possibly with some minor adaptive changes.
Designing an appropriate tokenomics and advancing decentralization for the trust.ai identity layer are vital for fostering trust and transparency within the ecosystem. By creating a decentralized trust model, we reduce reliance on a single entity, enhancing user confidence while incentivizing participation and contribution. This approach not only mitigates risks associated with single points of failure but also protects user rights, including data privacy. Ultimately, establishing a decentralized identity system is essential for adapting to future technological advancements and enhancing societal trust in AI and blockchain applications, aligning with the foundational principles of decentralization in the broader blockchain community.
