# Autonomous AI agents can’t be treated like human operators: InfStones’ Sili Zhao > Source: > Published: 2026-08-15 16:09:00+00:00 The convergence of blockchain technology and AI has given rise to 24X7 [AI agents](https://thecoinheadlines.com/crypto/coinfello-launches-fello-2-ai-agent-that-automates-defi-strategies-from-plain-language/article-29146/). Some of the biggest crypto platforms and infrastructure providers have rapidly rolled out production-ready support for autonomous AI agents, moving beyond simple chatbots into true onchain execution. Some recent rollouts include [OKX with its Agentic Wallet](https://thecoinheadlines.com/crypto/okx-unveils-ai-agent-payment-framework-amid-push-for-autonomous-transactions/article-15754/), Binance with modular agent skills, [Coinbase with its Coinbase for Agents,](https://thecoinheadlines.com/tech-and-ai/coinbase-launches-coinbase-for-agents-enabling-ai-tools-to-execute-crypto-transactions/article-22039/) and KuCoin with its Skills Hub. These AI agents, powered by blockchain technology, are executing live trading calls and moving money seamlessly across regions and time zones. This has led to calls for regulation and safeguards to ensure safety of funds and transparency. In an exclusive interview, Sili Zhao, Partner at InfStones talks about the impact of these AI agents on trading, the need for regulation, and the shift towards machine-to-machine blockchain infrastructure. ## Sili, we are seeing a massive paradigm shift where autonomous AI agents are moving from simple assistants to active economic actors capable of executing complex financial strategies on-chain. As these agents begin trading assets independently at machine speed, what internal technical and cryptographic safety guardrails must be hardcoded into the infrastructure to prevent rogue agents, market manipulation, or systemic capital drains? I think the biggest mistake would be treating an AI agent like a human operator and simply giving it a private key with unlimited authority. At machine speed, you have to assume an agent can make a bad decision, get compromised, or behave in ways you didn’t expect. The guardrails really need to sit at the infrastructure and wallet layer, outside of the AI itself. Things like spending limits, transaction and contract allowlists, rate limits, and automatic circuit breakers should be enforced before a transaction can actually be signed. Private keys also shouldn’t be directly exposed to the agent. This is also why we’ve been exploring concepts like Abstract Wallet at InfStones. The idea is to separate what an agent wants to do from what it is actually authorized to do on-chain. You can give AI enough flexibility to operate autonomously, while still putting hard cryptographic boundaries around its access to capital. For me, it really comes down to limiting the blast radius. You don’t assume the AI will never fail — you build the infrastructure so that when something does go wrong, one agent or one compromised credential can’t drain the entire system. ## A core theme of this shifting landscape is how regulated digital-asset ecosystems are emerging globally, largely propelled by fiat-backed stablecoins. From your perspective as an industry expert, how are strict regulatory frameworks—like Europe’s MiCA or the evolving frameworks in the UAE and the US—actually accelerating stablecoin adoption among mainstream enterprises, rather than hindering it? From what we’ve seen working with clients, one of the biggest barriers to [stablecoin adoption ](https://thecoinheadlines.com/blockchain/stablecoins-ai-will-drive-uaes-expat-wealth-vault22s-stephen-ong/article-28706/)has actually been regulatory uncertainty. When the landscape is unclear, enterprises become hesitant — not because they don’t see the value of the technology, but because they’re not confident moving forward when they don’t fully understand the regulatory or compliance risk. I don’t see compliance as something designed to hinder adoption. Good regulation is there to protect clients and market participants from rogue actions and create a safer environment for everyone to operate in. When the rules around reserves, custody, licensing, and consumer protection are clearly defined, companies can make decisions with much more confidence. That’s why we welcome well-defined and clearly laid-out regulations. In many cases, that clarity is what allows companies to move from exploring stablecoins to actually deploying real use cases. ## If blockchain infrastructure increasingly transitions to support machine-to-machine transactions, the definition of a ‘crypto user’ fundamentally changes. When AI agents start dominating on-chain volume for everyday tasks, what will the global digital-asset ecosystem look like in three to five years, and how will traditional institutional players adapt to a world run by autonomous algorithmic entities? I think blockchain has a lot of potential to support the financial industry and eventually become as widely used as many of the technologies we rely on today. AI is transforming almost every part of our lives, and blockchain will be part of that transformation. Our industry is also working to make blockchain more seamless, from applications all the way down to infrastructure. Ideally, users shouldn’t even need to know that blockchain is powering the application — it should just work. We already have automated trading in traditional finance today. Blockchain can make that automation more transparent and verifiable, whether it’s on an enterprise chain where privacy matters, or a public chain where transparency is more valued. ## As global jurisdictions deploy highly localized and often conflicting compliance standards for digital assets, scaling infrastructure remains a friction point. How can global web3 protocols maintain a borderless, decentralized architecture while simultaneously adhering to strict localized data residency laws, KYC/AML mandates, and sovereign security protocols? We’re already seeing this with clients today. Depending on the market, we support localized requirements around where infrastructure is deployed, where data is stored, and how access and security are managed, while still allowing clients to participate in the broader blockchain ecosystem. In some cases, even the protocol itself can be localized to serve a specific country or market, while remaining interoperable with other regions. The interaction between those ecosystems can then be governed according to the requirements of the relevant regulators. I think that balance between localization and interoperability will become increasingly important as blockchain adoption grows globally. ## You have navigated the crypto space through multiple severe market cycles and downturns, watching the industry evolve from a speculative playground into a structured, institutional asset class. Looking at the current intersection of AI, blockchain, and global regulation, what is the most critical macroeconomic impact this mature infrastructure will have on the broader legacy financial system? Something that needs to completely disrupt the financial system as we know it today. It can make moving money and assets faster, more transparent, more accessible globally, and less dependent on multiple intermediaries. We’ve even seen organizations like the UN explore how blockchain can help bring financial services to unbanked and underserved communities. To me, that’s one of the more exciting possibilities — not just making financial services more accessible, but making them more economical to access for everyone. There are many things in traditional finance that already work very well. Blockchain doesn’t need to replace them. If we can use it to reduce cost, improve settlement, expand access, and connect financial systems more efficiently, that’s already a very meaningful impact.