How Developers Are Solving AI’s Financial Limitations with Blockchain

One of many limitations of AI, and by extension robots, is their incapability to handle transactions autonomously. To handle this, builders are actually integrating AI with blockchain infrastructure, permitting AI brokers to transact securely, handle digital belongings, and execute sensible contracts.

At ETH Denver 2025, Decrypt spoke with trade specialists from the Coinbase Developer Platform, AI and robotics builders OpenMind and Robonomics to discover how AI-powered humanoid robots and different methods leverage blockchain to operate as unbiased financial brokers.

“Out of the field, AI doesn’t work together with the blockchain in any respect,” Lead Developer at Coinbase Developer Platform Nemil Dalal advised Decrypt. “If you would like it to have a pockets, ship cash, or switch funds to totally different locations, it will possibly’t try this. It can also’t join a checking account by itself. This can be a main limitation as a result of AI, in lots of circumstances, wants the flexibility to transact.”

Recognizing this hole, the Coinbase Developer Platform developed a framework known as Agent Package, which the corporate launched in November, Dalal stated. Agent Package permits AI to be given a pockets, enabling it to carry out any on-chain motion {that a} human can.

Calling it one of many earliest use circumstances for AI brokers, Dalal famous the potential of AI to deal with duties akin to monitoring the web and conducting enterprise on behalf of customers.

“To date, over 2,000 builders have constructed on prime of Agent Package, contributing to over $100 million in worth,” he stated. “Some of the common use circumstances is monitoring one thing on-chain and executing an motion in response—primarily an ‘if this, then try this’ automation.”

As AI’s affect grows throughout a number of industries, blockchain builders are more and more exploring its potential for crypto merchants and traders. Nonetheless, Dalal acknowledged AI hallucinations stay an issue within the trade.

“It might take actions or present info that’s not true,” he stated. “When actual cash is concerned, this difficulty turns into much more problematic.”

Dalal stated mechanisms that give customers larger management and oversight are wanted to make sure AI can safely deal with monetary transactions with out risking unintended actions.

“For example, if a transaction is under a sure greenback quantity, the AI can proceed routinely. If it exceeds that threshold, approval by the consumer is perhaps required,” he stated. “Equally, if the AI is paying a brand new occasion for the primary time, it could have to whitelist them, whereas subsequent transactions wouldn’t require further approval.”

Regardless of these considerations, Jan Liphardt, founding father of open-source AI and robotics startup OpenMind, stated AI and robotics present a singular and compelling use case for crypto.

“Blockchains had been constructed for machines, they usually’re a vital a part of the infrastructure that is wanted for machines to work nicely with folks,” Liphardt advised Decrypt.

A professor of bioengineering at Stanford College, Liphardt highlighted the challenges AI-driven humanoid robots face in human-centric environments, akin to banks and airways, which stay unprepared to accommodate them.

“We’re seeing unbelievable quantities of friction as considering machines attempt to navigate the human world, and an enormous a part of that could be a lack of id,” he stated. To mitigate this friction, OpenMind designed Iris to make use of a number of cryptocurrency wallets to facilitate transactions.

“She has two wallets,” Liphardt stated. “One is a straight Ethereum pockets, and the opposite is a Coinbase pockets,” he stated, including that Iris holds the personal keys to the wallets, checking the balances each six seconds.

Balancing innovation with security

As robots change into a part of human methods, questions emerge about their id, how they earn and handle cash, how they work together with folks, Liphardt stated.

What guidelines govern them and the way these guidelines could be modified are additionally entrance of thoughts, with Liphardt including decentralized ledgers are well-suited to deal with challenges as they already assist lots of these use circumstances.

Along with managing interplay with the true world, builders have turned to blockchain expertise to implement clear and immutable governance guidelines to deal with considerations about humanoid robots.

“That is why we write governance and habits guidelines onto Ethereum,” Liphardt defined. “When a humanoid robotic boots up, it reads these guidelines. If somebody asks, ‘Why ought to we belief this pc?’ you possibly can merely level them to the governance guidelines—they’re immutable and publicly accessible. Basically, we tailored Asimov’s Three Legal guidelines of Robotics and encoded them onto Ethereum.”

Whereas this offers some safeguards, Liphardt acknowledged it’s not an entire answer.

To alleviate worry of humanoid robots, influenced by movies like “The Terminator,” Liphardt famous that medical robots are smaller and extra approachable than these utilized in warehouses or hazardous settings.

For Vitaly Bulatov, evangelist with robotics startup Robonomics, the developer of the humanoid robotic Unitree G1, blockchain expertise provides sources that smaller robotics builders could not have entry to.

“Whereas we are able to run small LLMs on the robotic itself, we’ll possible want extra compute to totally allow this cognitive layer,” Bulatov advised Decrypt. “In our case, we’re constructed on Polkadot, so we function a gateway to Polkadot itself. We use compute suppliers on Polkadot to realize further computational energy for this cognitive layer.”

Bulatov stated that to combine autonomous robots seamlessly into our economic system, they have to be unbiased brokers able to managing transactions and contractual relationships.

“What we do is actually give these robots an id, a pockets, and the flexibility to enter contractual obligations with folks,” he stated.

The system captures information—together with error messages and video feeds—every time a cost is processed. Based on Bulatov, this info is recorded with a decentralized storage supplier on Polkadot, permitting for verification of activity completion.

“This fashion, irrespective of how massive the autonomous system turns into—whether or not it’s hundreds of thousands of robots or a complete sensible metropolis—we are able to monitor that, based mostly on cost, the technical execution occurs,” Bulatov stated.

Edited by Sebastian Sinclair