Ai Agents Bring New Rules For Crypto Wallets

What Happened

This creates a new product category around controlled autonomy. The user keeps ownership of the funds, while software handles repetitive execution under rules set in advance.

Payments Come First

Adrian Wall sees payments as the earliest major use case for AI agents, since payment mandates can be narrowed by amount, recipient, asset type, and timing.

Market Context

AI agents are entering crypto through wallets, exchanges, payment apps, trading systems, and portfolio tools. Once an agent receives signing authority, it can prepare transactions, rebalance assets, pay invoices, use smart contracts, and move across on-chain apps at software speed.

Stablecoins make cross-border payments a natural area for agent activity, especially in markets where bank transfers remain slow, expensive, or difficult to reconcile.

Trading and portfolio management are also ready from a technical view, but Wall placed more emphasis on governance than execution.

“Trading and portfolio management are technically mature enough today,” he said, adding the harder challenge is “whether authorization frameworks and loss limits are sophisticated enough to keep an agent’s mandate from drifting beyond what the user intended.”

Fernando Lillo Aranda said AI agents can improve automation, but users should give capital access gradually.

“AI agents can unlock automation, but capital access should always be progressive,” Lillo Aranda said.

Capital controls come first. Lillo Aranda said users should “cap maximum allocation, daily loss, position size, and withdrawal amounts.”

Permission controls come next. Users should “separate permissions for monitoring, trading, rebalancing, and fund movement,” he said.

Why It Matters

BeInCrypto spoke with Fernando Lillo Aranda, CMO at Zoomex; Federico Variola, CEO of Phemex; and Adrian Wall, Managing Director of the Digital Sovereignty Alliance, about early use cases, transaction approval, user limits, on-chain activity, and new risks once agents gain access to funds.

Identity may take longer, although Wall said decentralized identifiers and agent-assisted verification could reduce repeat authentication across fragmented digital services.

“The combination of decentralized identifiers and agent-driven verification is promising because it could reduce the burden on users who currently authenticate themselves repeatedly across fragmented systems,” Wall said.

Wallets were built around human review, while agents may prepare many actions across apps, contracts, and venues. Wall said wallet design now has to connect product choices with policy expectations.

“What we need is a tiered authorization model where the level of scrutiny matches the potential impact of the transaction,” Wall said.

This approach can separate monitoring, trade preparation, execution, and fund movement. A user may permit an agent to watch positions and draft trades, while reserving withdrawals and new contract access for manual approval.

Details

“Payments are the earliest use case because the parameters are well-defined and the mandate is constrained,” Wall said.

“Cross-border payments are especially compelling given the friction in legacy banking and the demonstrated efficiency of stablecoins,” Wall said.

Wallet Approvals Need Transaction-by-Transaction Controls

“The approval question is where policy and product design must converge, and it is where the industry has the most work left to do,” Wall said.

A strong approval model gives agents limited authority for routine actions while requiring human review for withdrawals, leverage, new contracts, and large swaps.

Fund Access Should Grow in Stages

He described the process as a gradual path from observation to assistance and execution. In practice, the agent first monitors and recommends, then prepares actions for approval, later receives limited execution rights, and eventually handles a larger mandate after reliable performance.