Ethereum has placed base layers to adjust autonomous agents, adjust the movement to place machines, and place them in machinery commercial on a direct route to chain settlements next year.
This month, the Ethereum Foundation established a dedicated DAI team with authority to promote agent identity, trust, and payments. This includes support for Agent 8004, including draft criteria for validation that entrench agent credentials and protocol level certification.
The initiative will constitute Ethereum as a settlement and coordinating tier for the agent economy with resistance to censorship and open access as core design goals, and the community draft around ERC-8004 outlines how on-chain identity and trust can enable automated systems to negotiate, post and execute bonds without binding.
The short-term deliverable is the progress of research and standards that can be adopted in wallets, middleware, and DAPP in 2026, creating a shared trust board for agent applications.
Token flow already reflects AI tilt in the crypto market.
AI-centric tokens such as Bittensor, Fetch.Ai (ASI), Internet computers, rendering, etc. maintain on-chain activity and relative price stability through Q3, outperforming the wider Altcoins during recent market drawdowns.
Koinly’s market roundup points to the ongoing demand for distributed calculation, inference, and agent frameworks, but the ecosystem report shows ICP push to render GPU marketplaces that render native app hosting and rendering GPU marketplaces that elicit stable usage from AI workloads.
For each token metric, the locked Defi Total Value was rebounded to the $100 billion region from around $72 billion in early 2025. This has helped invoke volume and charges that new AI-Native Defi Rails, such as Blackhole Dex, such as Avalanche, Sahara AI, and Moby AI, have maintained volumes. Token metrics place this in a wider rotation towards automated fluidity and agent execution that can operate across the chain via messaging and omnichin abstractions.
The payment stack converges on the agent’s use case at the protocol boundary. Google introduced the Payment (AP2) protocol to its agents in September, allowing software agents to request and confirm consumer payments via standardized flows. This was blocked through an interface with Crypto Settlement Rails, a building block for machine billing and subscription patterns.
According to Google Cloud, AP2 is designed around explicit user consent, verifiable agent identity, and reversible transactions for compliance, with early pilots including Ethereum and ICP integration through third-party connectors where Fiat accounts fill on-chain transfers.
As these pilots mature, wallets could treat agents as first-rate actors, and ERC-8004 style proof allows policies to spend caps per time frame, limit counterparties, and require human co-signations to high value thresholds.
The forward model now links plumbing upgrades to measurable network demand.
Token Metrics September Scenario Work Project AI Smart Agent will reach 15-20% of Defi transaction volume by quarter. This will place AI integration protocols in 2026 in a TVL range of $200-300 billion, if sustained and amplified by Ethereum’s Dai roadmap.
In the same analysis, feedthroughs are framed into the use of base layers. Agent Identity and Execution Agreement Gas Usage rose 30-40% over the quarter in 2026, and see widespread adoption beyond custody of standards like ERC-8004, consumer wallets, and DAO middleware.
In practice, this means governance, Treasury rebalance, fee routing, and cross-chain liquidity management can be performed by software agents operating with risk limits, insurance, and verifiable credentials on the chain.
The security outcome is another lever in the adoption curve. Adaptive’s academic and industry studies of AI-assisted contracts point to a sharp drop in successful exploits when contracts can detect abnormalities, tuning parameters, and suspicious flows of isolation in near real time.
Early models show a reduction of up to 70% in successful attacks of systems pairing rule-based controls with learning heuristics compared to static parameter schemes. This result relies on transparent update policies and monitorable on-chain behavior to avoid the creation of opaque control surfaces.
The context of macros is beginning to change from concept to pilot.
US and European regulatory agendas include workstreams on automated financial agents, adaptation contract transparency, and model risk disclosures.
DLA Piper’s September brief and other legal trackers explain the paths that make agent identity, usage policies and exception handling easy to read for regulators and counterparties.
Recent enforcement themes focus on the effectiveness of control, rather than banning technology that supports agent-operated runways in compliance with standard maturation.
RecruitBlock remains supportive, with RecruitBlock recording a 22% increase in its role at the intersection of AI and blockchain in 2025 year-on-year, and is a critical pipeline when protocol engineers, data infrastructure, and applied encryption, agent frameworks reach production scale beyond consumer and enterprise touchpoints.
Cross-market, machine economy lenses are not limited to a single stack. Avalanche hosts AI Gaban liquidity via Blackhole Dex, while Ethereum focuses on identity and payments, hosting and low latency inference for nearby chain-on-chain apps, providing GPU resources for training and serving models.
Koinly and Token’s metric coverage places them in complementary roles rather than directly alternatives, as the paper expands on the growing demand for distributed inference and market adjustment, and agents become default actors for payment, fulfillment, and protocol operation.
If the ICP growth model for native AI hosting is maintained, the on-chain inference cycle could reduce latency by half by 2026, making agent interactivity feasible in user-facing applications such as intent torquers, real-time hedging, supply chain or IoT payments.
| protocol | Primary AI functions | On-Chain Volume or TVL, September 2025 | Forward focus |
|---|---|---|---|
| Ethereum | Agent Identity and Payment, ERC-8004, DAI Team | $38B+ | Agent Trust and Coordination Layer |
| Bitenser, Tao | Decentralized Training and Inference Market | EST $1.4 billion. | Open AI calculation exchange |
| fetch.ai, fet | Autonomous Economic Agent, DAPP Infrastructure | $640 million EST. | Machine-to-machine adjustment |
| Render, RNDR | Distributed GPUs and Inference | ~$985m | Calculate the backbone of on-chain AI |
| Internet computers, ICPs | Native on-chain AI app hosting | 800 million dollars+ | Low agent dap latency |
| Black holedex, avalanche | AI-Gabenized AMMs and Liquidity | $193 million | Unauthorized Agent Transactions |
The scenarios are divided into three buckets.
The base case integrates Ethereum in the base case as at least a quarter of new DAPPS have adopted agent automation by 2026 and agent automation based on proof of programmable policies that employ governance, finance, fees and payments.
Bullpass turns on a full-machine economy where agents handle bilateral negotiations and fulfillment across consumer and corporate contexts, with DEFI TVL exceeding $300 billion, and decentralized AI API marketplaces reaching the critical mass of long-term tail services.
The bear case focuses on regulatory licensing of agents and ongoing centralization of computing and model access.
DLA Piper overview and policy trackers point to transparency and control standards as fulcrums rather than a complete ban, but computing centralization remains known constraints.
Investors and builders move from token narratives to measurable recruitment triggers.
On the standard side, the ERC-8004 is a core watch item. This is because wallets and custody providers need to implement agent proof checks, recovery flows, and policy enforcement to operate safely in consumer situations.
On the payer, AP2 pilots, when extended to large crypto rails, provide initial reproducibility patterns for subscriptions, usage claims, and fulfillment between non-human actors, exposes fine restrictions and approvals by putting pressure on bridges and account abstraction stacks.
In terms of security, adaptive control reduces losses, unlocks more autonomous governance, particularly for parameter tuning, in the field. Each of these tracks has public milestones that you can monitor without relying solely on price charts.
The open questions go beyond whether the agent trades. It is where settlement and trust checks occur.
If identity, proof, and policy live on the chain, the machine economy defaults to public ledgers and Defi becomes the operating system of non-human economic activity. If these checks remain on closed platforms, the role of Crypto will collapse into the bridge and payment rails.
With Ethereum’s DAI mandate, the AP2 pathway for agent payments, and the measurable shift in developers being adopted towards the role of AI X Crypto, the centre of gravity is moving towards verifiable on-chain adjustments that treat agents as first-class participants in the market.
It is mentioned in this article
(tagstotranslate)Ethereum
