Agent-based AI may reshape the way users interact with cryptocurrency wallets in the future—especially in terms of transactions and payments. While AI and blockchain executives point out that this could be secure, it also comes with new risks.
Last week, cryptocurrency exchange Coinbase announced its new tool, Payments MCP, which provides AI agents access to the same on-chain financial tools that people use.
Announcing Payments MCP, the easiest way for AI agents to get on-chain via x402. 🚀 It lets LLM models like Claude, Gemini, and ChatGPT gain access to on-chain tools like wallets, onramp, and payments with no API key required. 🧵 pic.twitter.com/MSnIaecx0O
The Coinbase developer platform stated in a release that when this tool is paired with large language models like Claude, Gemini, and Codex, they can access cryptocurrency wallets and autonomously make payments.
According to the Coinbase developer platform, AI agents powered by Payments MCP can make payments, compute, retrieve paywall data, tip creators, and manage certain business operations via the x402 protocol. x402 is an open, native web payment protocol that facilitates instant stablecoin payments.
"This marks a new phase of agent commerce, where AI agents can play a role in the global economy," the Coinbase developer platform stated.
Aaron Ratcliff, head of attribution at blockchain intelligence firm Merkle Science, told Cointelegraph that from a security perspective, allowing AI agents access to your wallet adds a layer of trust to a system designed to be trustless.
If the system is built correctly, it can be secure, but Ratcliff believes that "security" ultimately depends on the cryptocurrency user.
A survey conducted in April by cryptocurrency data aggregator CoinGecko of 2,632 cryptocurrency users found that most users are comfortable with AI trading on their behalf; 87% of users indicated they would allow AI agents to manage at least one-tenth of their cryptocurrency portfolio.
Ratcliff stated that if AI is used in a portfolio, malicious actors could exploit certain security risks. Prompt or instruction injection could allow someone to hijack the system.
Man-in-the-middle attacks—where hackers insert themselves between entities in a communication channel to steal data—could also redirect transactions.
"AI could also interact with scam tokens, miss honeypots or rug pulls, or handle slippage so poorly that it burns user funds," Ratcliff added.
At the same time, Ratcliff believes that compliance gaps could lead to issues, such as a lack of controls to prevent AI from sending funds to sanctioned addresses or exchanges.
Sean Ren, co-founder of AI-native blockchain platform Sahara AI, told Cointelegraph in an interview that for Coinbase, the exchange's tools leverage model context protocols, "which is the gold standard for security when set up correctly."
"They essentially act as gatekeepers between the AI model and your wallet. Agents can only perform specific, approved actions—like checking balances or preparing payments for your confirmation—rather than freely transferring funds or changing wallet settings," he said.
"These actions are designed to be restricted, so even if someone tries to deceive the AI through prompt injection or similar methods, it cannot complete transactions on its own," Ren added.
However, Ren also stated that being more secure does not mean foolproof, and users still need to pay attention to the actions of AI agents regarding their portfolios.
Brian Huang, co-founder and CEO of AI-driven cryptocurrency portfolio management platform Glider, told Cointelegraph that basic functions like sending, swapping, and lending are a good starting point for agents, but the field is still in its early stages.
"These are simple operations that can be completed with a click—you wouldn't ask ChatGPT to transfer money to a friend, right? Many of these operations take longer through agents," he said.
Huang predicts that more complex operations, such as portfolio management, rebalancing, and personalized financial advice, may follow and become more effective use cases.
"The customization that agents can provide, the number of variables they can consider, far exceeds what any human can offer," he said.
Related: IBM's "Digital Asset Harbor" aims to transform cryptocurrency into enterprise infrastructure
Original article: “AI Agents Want to Manage Your Crypto Wallet—But Is It Safe?”
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
