The Ethereum mainnet is set to implement the ERC-8004 upgrade, enabling AI agents to interact with organizations.
ERC-8004 is going live on mainnet soon.
— Ethereum (@ethereum) January 27, 2026
By enabling discovery and portable reputation, ERC-8004 allows AI agents to interact across organizations ensuring credibility travels everywhere.
This unlocks a global market where AI services can interoperate without gatekeepers. https://t.co/Yrl0rvnSxj
This innovation will enable the discovery of digital assistants and their services without a central structure. They will also gain a reputation that is not tied to a single platform.
ERC-8004 allows verification of:
- the identity of the AI agent;
- its past activities;
- whether it can be trusted.
These changes will open a global market for AI-based services, allowing assistants to interact with one another without intermediaries.
Under the proposed ERC-8004 from August 2025, each agent will receive an identifier in a special registry contract. This ensures compatibility with the current infrastructure and allows asset transfers between owners.
Trust models for AI agents are categorized by security levels, such as ordering pizza (low risk) or medical diagnosis (high risk).
The official activation date for the upgrade has not been announced. However, Marco De Rossi, co-author of the standard and head of AI at MetaMask, pointed to January 29, 2026.
“Ethereum occupies a unique position as a platform that ensures security and calculations in interactions between artificial intelligences,” said Davide Crapis, head of the AI department at the Ethereum Foundation.
Identity, Reputation, Validation
ERC-8004 introduces a system of three registries for verifying and enhancing trust in AI agents. The solution is compatible with both the mainnet and L2 protocols.
The identity registry assigns each agent a portable, censorship-resistant identifier. Compatibility with NFT standards allows easy transfer of the assistant's profile between different applications.
The reputation registry provides an interface for signed reviews from clients, forming a sort of rating for digital assistants.
The validation registry is responsible for auditing: agents request verification of their actions, and validators record the results on the blockchain.
The Future of Payments
The crypto community is actively discussing a future where AI agents manage finances: conducting transactions, purchasing goods, and operating without human involvement.
Ethereum co-founder Vitalik Buterin noted that in the next 5–10 years, artificial intelligence will start processing “a multitude of micropayments.” However, full automation for larger sums is too risky.
“AI is easy to attack. But it is suitable for small payments. AI can also serve as a risk management tool for large transfers—checking if a $100,000 transaction looks suspicious or not. I think we will gradually discover the right ways to combine AI with payments. There will be successes, but mistakes will also occur,” he explained.
Paxos identified AI agents as a potential driver for the stablecoin sector. According to co-founder Bhav Kotecha, fragmentation in the market across jurisdictions and various “pegged” token models can be addressed with digital assistants.
Algorithms can instantly switch between different platforms and assets, selecting the most favorable conditions for transactions.
“This means that fragmentation is not necessarily a limiting factor—it can become an optimizer at the market level, where artificial intelligence provides liquidity flows to the most efficient issuers. Over time, this could lead to lower fees and compel providers to compete based on fundamental metrics,” Kotecha stated.
Galaxy Digital CEO Mike Novogratz expects that “in the not-too-distant future, the largest users of stablecoins will be neural networks.” Users will be able to delegate daily purchases and bill payments to AI agents, which will handle transactions.
Investor and Shark Tank star Kevin O’Leary remarked that artificial intelligence and blockchain will trigger “the next revolution in business.” AI will be able to make retail purchases independently, while distributed ledger technology will process payments.
Solutions for the AI economy of the future are being developed.
The largest cryptocurrency exchange in the U.S., Coinbase, launched the Payments MCP system, allowing language models like Claude from Anthropic and Gemini from Google to access cryptocurrency wallets, fiat gateways, and payments.
PayPal will implement the Agentic Commerce Protocol (ACP) and the Instant Checkout feature from OpenAI, allowing users to make purchases through ChatGPT.
ACP is an open standard that enables sellers to list their products in AI applications, allowing users to make purchases through agents.
The Instant Checkout feature, launched in September, allows users to confirm orders, delivery, and payment without leaving ChatGPT.
At the end of October, the Quack AI team introduced x402 BNB—“a unified layer for signatures, payments, and management in the agent economy on BNB Chain.” This tool will replace multiple steps with a single signature: confirmation → transfer → payment.
x402 BNB “eliminates the problem at the protocol level,” providing “gasless one-click transactions that can be verified, audited, and controlled.”
Ethereum Not Profiting
Fees on the Ethereum network have reached their lowest level since May 2017.
Source: Glassnode.However, this is not due to a decline in activity. In January, the second-largest cryptocurrency network set a historical record for the number of transfers. The seven-day average of transactions approached 2.5 million—double last year's figures.
The sharp increase in on-chain activity began in mid-December after a prolonged decline. Recent upgrades have ensured the network's scalability.
- Activation of the Fusaka upgrade in early winter;
- Fork of the Blob Parameter-Only (January 7), which increased the limit of BLOB objects in a block from 15 to 21.
Recall that in January, Glassnode recorded a twofold increase in the retention rate of active users on the Ethereum network.