Crypto AI Agent Requires User Approval for Every Trade

A new AI trading agent is challenging the high-speed execution norms in the cryptocurrency market by requiring explicit user approval for every transaction before it is executed. Developed by San Francisco-based OpenLedger, the agent aims to reintroduce human oversight into automated trading, mitigating risks associated with uncontrolled algorithmic behavior [1]. Unlike most AI crypto tools, which operate without human intervention, this agent will pause for user confirmation, ensuring that decisions are transparent and fully traceable [1].

The system operates using a “Proof of Attribution” framework, which links every suggested trade to its data sources and model versions. This allows users to audit the logic behind each recommendation. If the data trail is incomplete, the agent simply won’t proceed. Additionally, all prompts and context remain on the device by default, ensuring that private keys and transaction histories are not transmitted externally [1]. OpenLedger emphasizes privacy by storing no user data in the cloud unless explicitly needed for functionality, and even then, such data is anonymized and not linked to individual wallets or identities [1].

The rollout of the AI agent within Trust Wallet is set to begin in October, initially limited to a few hundred users, with wider access planned for early next year. OpenLedger’s approach aligns with growing concerns about the risks of AI-driven collusion and “hallucinations” in trading systems. In May, researchers from Princeton University demonstrated how memory-based manipulation attacks could trick crypto AI agents into approving unauthorized transactions, highlighting the need for robust verification mechanisms [1].

This development comes amid a broader surge in AI-powered tools in the crypto space. The sector’s market cap for AI-related projects reached $5.7 billion in July, as developers integrate AI into decentralized finance (DeFi) platforms. Other major players are also entering the space; for example, Coinbase’s rebranded wallet app, Base, announced an AI feature in July that allows users to trade and send crypto using natural language prompts [1].

While OpenLedger is not the only company exploring AI for crypto wallets, its focus on transparency and user control sets it apart. The company’s core contributor, Kamesh, emphasized that the AI acts more like a “smart co-pilot” than an autonomous decision-maker. Users retain full authority over trades, with AI serving as a tool to provide insights and suggestions. “You’re still flying the plane,” Kamesh said [1].

The introduction of user-confirmation-based AI trading reflects a shift in the industry toward slower, more deliberate execution. This is in contrast to the current trend of rapid AI deployment, where automation often outpaces regulatory understanding and control. The balance between speed and accountability remains a central issue in the ongoing evolution of AI-driven trading [1].

Source:

[1] title: In a Market Obsessed With Speed, This Crypto AI Agent Wants You to Slow Down, url: https://decrypt.co/333417/market-obsessed-speed-crypto-ai-agent-wants-slow-down/

[2] title: 'Artificial stupidity' made AI trading bots spontaneously form ..., url: https://www.aol.com/finance/artificial-stupidity-made-ai-trading-110500181.html

[3] title: Tokens are getting more expensive, url: https://x.com/Cryptotrissy/status/1952306827963871542