Account Abstraction in 2026: One Account, Too Many Chains

Open your wallet app and it shows one address, the same string of characters whether you tap Ethereum, Base, Arbitrum or Optimism. It feels like one account. It is not. Your USDC might sit on Base, an NFT on Arbitrum, a game item on Ronin, and the smart-account features you switched on last month may only exist on the single chain where you turned them on. Account abstraction, the technology that turned a crypto wallet from a bare key into programmable software, is the biggest wallet story of the past two years. It made your account smart. It did not make it portable.

The plumbing is no longer niche. As of early October 2026, ETH trades around $2,685 (CoinDesk), and BundleBear counts more than 1.3 billion UserOperations across over 68 million ERC-4337 accounts (BundleBear); EIP-7702, the upgrade that lets an ordinary wallet borrow smart-account powers, shows north of 63 million live delegations (BundleBear). Three days from now, on 6 October, Ethereum rehearses its next upgrade, Glamsterdam, on the Sepolia testnet. And just weeks ago, the two teams pushing hardest to build account abstraction into the protocol itself, the Ethereum Foundation and Coinbase’s Base, publicly gave up on a shared design. Every one of those threads loops back to the same unsolved problem: a smart account that only understands one chain at a time.

This is a guide to account abstraction as it actually works in late 2026, built around the question most explainers skip: if my account is so smart, why can’t it follow me across chains? We will cover what a smart account is, the two ways you get one, the machinery that makes gasless transactions possible, and then the portability gap and the three separate efforts to close it.

What account abstraction actually means

An ordinary Ethereum wallet is an externally owned account, or EOA. Strip away the app and it is just a keypair: one secret key that signs, one public address that receives. The design is simple and it has one brutal rule. The key is the account. Lose it and the funds are gone; leak it and they are stolen; there is no reset, no account manager, no second factor baked in. An EOA cannot batch two actions into one, cannot set a spending limit, cannot let a friend help you recover access, and cannot pay its gas in anything but ETH.

Account abstraction is the idea of replacing that rigid keypair with a smart contract you control, so the rules of the account become programmable. Instead of one key with total power, a smart account can say: require two of my three keys for large transfers; let this game move my items but nothing else for the next hour; pay gas in USDC; and if I lose my phone, let my chosen guardians rotate the key after a delay. The concept has been discussed since 2016. What changed is that it finally shipped in forms ordinary users touch.

The important shift is this: in account abstraction, the account is software, not a secret. Verification (who is allowed to act) is separated from execution (what the account does), and both become code you can change. That single move is what enables passkeys, gasless transactions, session keys and social recovery. It is also what creates the portability problem, because that code has to live somewhere, and somewhere means a specific chain.

The two roads to a smart account

In 2026 there are two mainstream ways to end up with a smart account, and the difference matters for everything that follows.

The first is ERC-4337, finalized in March 2023. It builds account abstraction entirely on top of Ethereum, with no change to the protocol. Your account is a dedicated smart contract. Instead of a normal transaction you sign a message called a UserOperation, which travels through a separate mempool, gets packaged by an operator called a bundler, and executes through one shared contract called the EntryPoint. Because it needs no consensus change, it runs on Ethereum and nearly every EVM Layer 2 today.

The second is EIP-7702, which went live with the Pectra upgrade on 7 May 2025. Rather than make you migrate to a new contract account, 7702 lets your existing EOA temporarily point to smart-contract code through a new transaction type. You keep the same address and the same key, but the account can suddenly do smart-account things: batch actions, accept a gas sponsor, run session keys. Ambire shipped the first 7702 wallet; MetaMask, Bitget and others followed. The latest EntryPoint release (version 0.8, 2025) even wires the two approaches together, so a 7702 account can use the 4337 machinery (GitHub).

Marius van der Wijden, an Ethereum core developer, described 7702 as a change that “adds a new transaction type that allows existing wallets to emulate the functions of Account Abstraction wallets,” while cautioning that the ecosystem still needs to “evaluate all the rough edges” (DL News). Both roads lead to a programmable account; they differ in whether you move to a new contract or upgrade the account you already have.

The last row is the one this article is about: both smart-account models are per-chain.

The machinery behind a gasless transaction

Gasless is the feature new users notice first, and it is worth seeing what happens underneath, because the plumbing is run by companies, not by magic.

When you tap send on an ERC-4337 account, three roles come into play. A bundler is the operator that collects UserOperations, packs them into a regular transaction, and pays the base-layer gas to get them on chain, then is reimbursed inside the batch. The EntryPoint is the single audited contract every 4337 account routes through, which makes it a shared, security-critical chokepoint. And a paymaster is an optional contract that agrees to cover your gas, either because an app is subsidizing you or because you are paying in a token such as USDC instead of ETH.

That economy is real money. BundleBear shows paymasters have now sponsored more than $14.5 million in gas across ERC-4337 (BundleBear). Circle’s own paymaster lets any 4337 account pay gas in USDC on chains including Arbitrum and Base, and since 1 July 2025 it adds roughly a 10% surcharge on the gas it fronts (Circle). Gasless never meant free; it meant someone else pays first and usually bills you later.

It also means concentration. A small number of operators handle most of the traffic: through 2026, Pimlico and Alchemy together have processed the majority of all UserOperations (BundleBear). The convenience of never touching gas is paid for with a dependence on a handful of off-chain services, the same pattern that makes cross-chain infrastructure so fragile.

What a smart account actually lets you do

Before the catch, it helps to be concrete about why anyone bothers. A smart account can do several things a bare key never could.

• Batch actions: approve a token and swap it in a single click, instead of two signatures and two gas payments.
• Sponsor gas: let an app pay your fees, or pay them yourself in a stablecoin, so a first-time user never has to buy ETH just to move.
• Run session keys: grant a scoped, time-limited permission so a game or app can act for you (for example, play for one hour, or spend up to 50 USDC a day) without a pop-up on every move. This is the reason blockchain games finally feel like games rather than a signature-prompt simulator.
• Replace the seed phrase with a passkey: the same Face ID or fingerprint credential you already use, made cheap on chain by the RIP-7212 precompile, which cut the cost of an on-chain passkey check from roughly 300,000 gas to about 3,450 (Alchemy). The FIDO Alliance counts around 5 billion passkeys in use worldwide (FIDO Alliance).
• Offer social recovery: nominate guardians or a backup device so a lost key is a recoverable inconvenience, not a funeral.

Vitalik Buterin has argued for years that usability and self-custody are not opposites, calling the human brain “an ASIC for keeping track of relationships with other people,” the reasoning behind social-recovery wallets (vitalik.eth.limo). Every one of those features, though, is a property of a specific contract on a specific chain. Turn them on where? That is the catch.

The catch: your smart account lives on one chain

Here is the part most account-abstraction explainers wave past. A smart account is code, and code is deployed to one chain at a time.

If you use ERC-4337, your account is a contract. Deploying it on Base does not deploy it on Arbitrum; those are two separate contracts that happen to be controlled by the same key. You can easily end up with your smart account activated on one network and simply absent on another.

If you use EIP-7702, the situation is subtler and arguably worse, because it looks like it should just work. Your EOA address is identical on every EVM chain, a property of how keys derive. But the 7702 delegation, the pointer from your address to the contract code, is authorized per chain. The authorization you sign is a tuple of (chain id, contract address, nonce), and a non-zero chain id makes it valid only on that chain (EIP-7702 spec). Delegate on Ethereum and your address on Optimism is still a plain EOA until you delegate there too.

The spec does allow a shortcut: set the chain id to zero and the authorization is valid on any chain. That sounds convenient, and it is a footgun. The same contract address can hold different code on different chains, so a chain-zero delegation can point your account at something you never intended elsewhere. That is why the standard tells wallets to default to per-chain authorizations and to warn loudly before anyone signs a chain-zero one (EIP-7702 spec). Convenience and safety pull in opposite directions, and portability loses.

Why one address is not one account

The confusion is understandable, because the address really is the same everywhere. Two things, though, are not shared across chains: your account’s code and your account’s state.

Think of the address as a street address that exists in a hundred parallel cities. The building at that address might be a house in one city, a vault in another, and an empty lot in a third. Your balances live in those buildings, not in the address itself. So you can hold 500 USDC on Base and 500 on Arbitrum and own, in a real sense, two separate piles that your wallet lists under one name. Spending the Base pile for something priced on Arbitrum takes a bridge or a swap, not a single click, and that is before anything smart is involved.

This is also where people lose money in ways that have nothing to do with hacking. Send funds to your own address on the wrong chain and they are not gone, but retrieving them means having the right account set up on that chain. Deposit to an exchange over a network it does not credit and a support ticket follows. The cross-chain fragility smart accounts inherit is the same fragility that makes bridges the industry’s most dangerous plumbing, a theme we have traced through where cross-chain security keeps breaking.

Chain abstraction: making many chains feel like one

The industry’s first answer to all this is chain abstraction: hide the chains from the user entirely. The goal, as its builders describe it, is that you never have to pick, see or think about which chain you are on (Eco). You see one balance and one button, and the complexity is pushed down into infrastructure.

Under the hood it usually has three layers. There is an account layer, a single smart account (ERC-4337 or an EIP-7702 EOA) that represents you. There is an execution layer built on intents: instead of signing a specific transaction, you sign what you want (pay this merchant 20 USDC), expressed through emerging standards such as ERC-7683, and a competitive network of solvers works out how to make it true across chains. And there is a liquidity layer, where market makers and protocols such as Across and CoW Protocol pre-position inventory so a solver can deliver funds on the destination chain instantly and settle up behind the scenes.

The clearest example of the experience is the unified balance: hold 500 USDC on Arbitrum and 500 on Base, and a chain-abstracted wallet shows you a single 1,000 USDC that you can spend anywhere (Eco). Networks built around cross-chain intents have already moved billions of dollars in volume, the first real sign the pattern works outside demos. The honest caveat is that this is convenience layered on top of the same bridges and solvers, so it inherits their trust assumptions and their failure modes. It makes the chains invisible; it does not make them one.

Keystore rollups: fixing portability at the root

Chain abstraction smooths the experience, but it leaves a deeper problem. If your account rules (which keys, which guardians, which recovery settings) live in a contract on one chain, then changing them means changing them everywhere, chain by chain. Rotate a compromised key and you have to repeat the operation on every network you use.

The cleaner fix is a keystore rollup: put the answer to who is allowed to control this account in one dedicated place, and let your accounts on every chain read from it. Vitalik Buterin sketched exactly this, a minimal rollup whose only job is to hold keystores, so a single update to your permissions propagates everywhere (Safe Foundation). Base has published its own keystore design (Base), and purpose-built keystore rollups have started to appear. Settings in one place, accounts everywhere: that is the end state chain abstraction reaches toward from the user side and keystore rollups from the protocol side.

Buterin frames the whole effort around a principle he calls intermediary minimization, that a good design should “maximize what you can do even if all the world’s infrastructure except the Ethereum chain itself goes down” (Cointelegraph). A keystore you control, rather than a bridge operator you trust, sits squarely in that spirit.

The fork in the road: EIP-8141 vs EIP-8130

If keystores and native support are the destination, 2026 is the year the road forked. The two teams who have pushed account abstraction hardest, the Ethereum Foundation and Coinbase’s Base, spent months trying to agree on a single way to build it into the protocol. In mid-September they announced they had given up on a shared standard (The Defiant).

Ethereum’s proposal is EIP-8141, called Frame Transactions. It splits a transaction into at least two parts: a VERIFY frame that checks the signature and authorizes the fee, and one or more EXECUTE frames that do the work, with built-in paymaster support and no reliance on the old ecrecover primitive. Its author list includes Vitalik Buterin, the Foundation has flagged it as a must-ship item for a future fork, and Buterin said in September that it “gets close to optimal.”

Base’s proposal is EIP-8130, built around an on-chain keystore: a new transaction type plus a registry of approved signers that names the authenticator per transaction. It is tuned for Base’s priorities of scale, customization and predictable costs, and 8130 accounts would still reach other chains through ERC-4337.

Derek Chiang, the ZeroDev founder now at Ethlabs who helped broker the talks, explained the breakup bluntly: “While we identified a number of technical solutions, they all required one side or the other to compromise at least a little bit on their core goals” (The Defiant). Read through this article’s lens, the split is not abstract: it is a disagreement about the cross-chain future of your account, and for now it hands the job of papering over the fragmentation to wallet developers.

Glamsterdam is a speed upgrade, not account abstraction

It is easy to assume the upgrade everyone is watching this week will deliver native accounts. It will not, and the distinction matters.

On 6 October 2026 at 13:53 UTC, Ethereum activates its Glamsterdam upgrade on the Sepolia testnet (CryptoTimes). Glamsterdam’s headline features are about throughput: enshrined proposer-builder separation (EIP-7732) and block-level access lists (EIP-7928) that let more execution run in parallel. It is Ethereum’s first coordinated step toward scaling the base layer, and a mainnet rollout is targeted loosely for the fourth quarter with no firm date; developers have cautioned that the upgrade has not yet run cleanly on a stable private devnet.

None of that is native account abstraction. The protocol-level AA work, EIP-8141’s frames, is slated for a later fork, not this one. So when you read that Glamsterdam is days away, the smart-account endgame is not. The testnet clock on your screen and the account-abstraction roadmap are two different calendars, and conflating them is the easiest mistake to make right now.

The security bill no one prices in

Programmability widens what an account can do, which also widens what can go wrong, and the cross-chain shape of smart accounts multiplies the cleanup.

Start with the delegations. On EIP-7702, more than 63 million live delegations sound like mass adoption, but the cumulative authorization count, past 260 million, is heavily inflated by automated sweeper bots that re-sign compromised accounts over and over (BundleBear). Security firm Wintermute found that more than 97% of early 7702 delegations pointed at identical sweeper code, a family nicknamed CrimeEnjoyor. Crucially, that is not a flaw in 7702: the accounts were already compromised, and the code is harmless to the rest of us (CoinDesk). But it shows how fast attackers moved in.

The more ordinary danger is blind signing: approving a transaction whose meaning your wallet cannot show you in plain language. It is how the industry’s biggest thefts happen, and it is why Ledger-originated clear-signing work (ERC-7730) was handed to the Ethereum Foundation in 2026 as shared infrastructure, under the rule that what you see should be what you sign (Ethereum Foundation). There is good news: phishing losses are falling. Scam Sniffer logged a roughly 83% drop in wallet-drainer losses in 2025, to around $84 million (Cointelegraph), even as a handful of 7702-specific scams emerged. The social-engineering threat has not vanished so much as moved; the organized, forced-labor scam industry we documented in pig butchering now drives more losses than code bugs do.

Then the portability tax reappears. A delegation you want to revoke must be revoked on each chain where you set one, because clearing it in one place does nothing elsewhere. Smart accounts made the good things repeatable across chains; they made the cleanup repeatable too.

Exchanges, custody and the SEC

Smart accounts complicate the one address exchanges rely on most: your deposit address. Because a 7702 address can carry delegated code, exchanges now screen deposits, checking whether an incoming address has a delegation (a 23-byte marker beginning 0xef0100) and handling it accordingly, chain by chain. Several large custodians also pair multi-party computation key management with 7702, offering institutions smart-account features without ever assembling a single private key.

For US readers the regulatory line is clearer than it used to be. In April 2026, the SEC’s Division of Trading and Markets issued staff guidance that software merely enabling users to transact from their own wallets is not acting as a broker (CoinDesk). In plain terms, a self-custody smart-account wallet is generally software, not a regulated intermediary, while the custodial exchange that holds your keys is squarely in scope. That boundary matters as wallets grow exchange-like features and exchanges ship self-custody wallets; who is regulated turns on who holds the keys, a question we keep returning to in coverage of crypto’s missing referee. The same convergence shows up in trading, where self-custody wallets now route orders to on-chain perpetuals venues, a shift we covered in perp DEXs going onshore.

For all the new features, the oldest argument in crypto still applies: self-custody is a responsibility, not only a convenience, which is the lesson at the center of Bitcoin’s own self-custody shakeout this year.

Which wallets actually do this in 2026

If you want to use a smart account today, the names to know fall into a few camps.

Safe remains the giant of contract accounts, especially for teams and treasuries. In the second quarter of 2026 it processed nearly 130 million transactions across more than 63 million accounts, held over $27 billion in assets, and by its own report handled close to one of every 48 dollars of stablecoin supply (The Block). Lukas Schor, who co-founded the Safe ecosystem, noted that “active accounts continued to rise, and the value transferred through Safe accounts on Ethereum, expressed in ETH, reached its quarterly high” even as prices fell (The Block).

For individuals, Base Account (the former Coinbase Smart Wallet) leans on passkeys and spend permissions; MetaMask Smart Accounts brought the 7702 path to a huge existing user base; Ambire shipped the first 7702 wallet; Ready (formerly Argent) carries the social-recovery torch; and Bitget markets cross-chain gas so you can spend without holding each chain’s native token. Modular standards sit under many of them, chiefly ERC-7579 (minimal, now the de facto choice for Safe, ZeroDev and Biconomy) and the heavier ERC-6900.

The portability question is the one to put to each of them: does this wallet activate my account on the chains I use, does it warn me about chain-zero delegations, and can it revoke per chain? In late 2026 the honest answer is partly, and inconsistently.

What it means for you right now

Account abstraction has genuinely arrived. Over 68 million ERC-4337 accounts and more than 63 million live 7702 delegations are not a pilot (BundleBear). The wins are real: no seed phrase, no scramble for gas, games you can actually play, recovery that does not end in an apology.

The honest caveat is that your account is still a per-chain object wearing a single address, and the three fixes (chain abstraction at the interface, keystore rollups at the protocol, native AA at the base layer) are all mid-flight, with the two native efforts now pulling in different directions. Until they land, a few habits pay off.

• Know which chain you are on, and which chains your smart account is actually set up on.
• Treat a valid-on-all-chains (chain-zero) delegation as a red flag, not a shortcut.
• If you revoke a delegation or rotate a key, do it on every chain where you set it.
• Do not confuse the Glamsterdam testnet clock with the account-abstraction roadmap; they are different calendars.
• Prefer wallets that show you, in plain language, exactly what you are signing.

The account became software. Making that software feel like one account, no matter how many chains it touches, is the work of the year ahead, not the year behind.

Frequently Asked Questions

What is account abstraction in simple terms?

Account abstraction replaces a crypto wallet’s rigid design, where one key gives total control, with a programmable smart contract. That lets your account batch actions, pay gas in tokens other than ETH, use a passkey instead of a seed phrase, set spending limits, and recover access if you lose a key. In short, your account becomes software whose rules you can change.

What is the difference between ERC-4337 and EIP-7702?

ERC-4337 gives you a brand-new smart-contract account and runs entirely on top of Ethereum, with no protocol change, using bundlers, paymasters and a shared EntryPoint contract. EIP-7702, live since the Pectra upgrade in May 2025, keeps your existing address and key but lets that account temporarily point to smart-contract code. One is a new account; the other upgrades the account you already have.

Why does my smart account not work across all chains?

Because a smart account is code, and code is deployed or delegated one chain at a time. Your address is the same on every EVM chain, but an ERC-4337 contract must be deployed on each chain separately, and an EIP-7702 delegation is authorized per chain. So your account can be smart on one network and a plain wallet on another until you set it up there too.

Is account abstraction safe to use?

The core standards are widely used and audited, and overall wallet-drainer losses fell about 83% in 2025. The real risks are blind signing (approving a transaction you cannot read in plain language), scam delegations from already-compromised keys, and forgetting to revoke permissions on every chain. Using wallets with clear signing and per-chain revocation, and treating chain-zero delegations as a red flag, removes most of the danger.

Will Ethereum’s Glamsterdam upgrade bring native account abstraction?

No. Glamsterdam, which reaches the Sepolia testnet on 6 October 2026, is a throughput upgrade (enshrined proposer-builder separation and block-level access lists), not account abstraction. Native, protocol-level account abstraction is a separate effort, and in 2026 Ethereum and Base split over how to build it, with Ethereum backing EIP-8141 and Base backing EIP-8130.

Yuki Tanaka covers wallets, exchanges and account abstraction for HOGE Wire.