Account Abstraction in 2026: The Endgame Keeps Slipping

Account abstraction has been the promised destination of Ethereum wallet design for close to a decade. The pitch is easy to state: your account should be programmable software, not a single private key that either works perfectly or loses everything at once. In 2026 that promise is half kept. Smart-account logic now runs on tens of millions of addresses, gasless transactions are ordinary, and passkeys are quietly replacing seed phrases in mainstream apps. Yet the version that was supposed to be the finish line, account abstraction built directly into the protocol, keeps sliding out of reach.

In the first week of September, Base pushed its native standard off a shipping upgrade and onto a fork with no date attached. It was the latest in a run of delays that has turned a supposed endgame into a moving target. This guide explains what account abstraction is, what already works today across wallets and exchanges, and why the native version everyone has been waiting for is still not here. Prices and figures below are current as of 14 September 2026, with ETH trading around $2,500 (CoinDesk).

What account abstraction actually means

Ethereum has always had two kinds of accounts. Externally owned accounts, or EOAs, are controlled by a private key; contract accounts are controlled by code. An EOA is what a seed phrase gives you in a wallet like MetaMask. It can hold assets and sign transactions, but it cannot do anything that a raw elliptic-curve signature cannot express. Account abstraction erases that hard boundary. It lets an account be governed by arbitrary logic: two-of-three signers, a daily spending cap, a passkey stored in your phone’s secure hardware, a time-locked recovery key, or a session key that can trade for an hour without prompting you on every move. The private key stops being the account itself and becomes just one of several ways to authorize it.

The problem this solves is the one that greets every newcomer. A seed phrase is a bearer instrument: whoever holds it owns everything, there is no reset button, and a large share of all Bitcoin ever mined is now considered permanently lost to forgotten keys, according to estimates compiled by Ledger. Gas has to be paid in the chain’s native token, so someone holding only stablecoins cannot move them. And every action is a separate signature, so approving a token and then spending it means two prompts and two chances to sign something malicious. Smart accounts were designed to fix all three at once, and after years of theory the tooling to do so is finally in production.

EOAs versus smart accounts, the core difference

There are now three account shapes a user might hold, and telling them apart matters because they behave differently at the wallet, the exchange, and the tax desk. A legacy EOA is the classic key-only account. An ERC-4337 smart account is a contract with its own address and its own rules. An EIP-7702 upgraded EOA is the hybrid that arrived in 2025: your existing key-based address, temporarily wearing smart-account clothes. The table below lays out how they compare.

The practical upshot: ERC-4337 gives you the fullest feature set but asks you to adopt a new address, while EIP-7702 upgrades the address you already have. Native account abstraction, when it eventually lands, aims to fold the best of both into the base protocol so that no side infrastructure is needed at all.

ERC-4337: account abstraction without changing Ethereum

ERC-4337, finalized in March 2023, delivered smart accounts without a hard fork (EIP-4337). Instead of ordinary transactions, users sign objects called UserOperations that live in a separate mempool. Operators known as bundlers pick them up, wrap them into a real transaction, and send them to a single audited contract called the EntryPoint, which verifies each operation against the account’s own logic before executing it. Paymasters, optional contracts that agree to cover gas, sit alongside. The current production EntryPoint is v0.7, and v0.8, released in 2025, added native support for the EIP-7702 flow (eth-infinitism releases).

Adoption has become substantial. By mid-September 2026 the standard had processed more than 1.28 billion UserOperations across roughly 66.5 million accounts with activity, with paymasters covering about $14.1 million in gas, according to BundleBear. The catch is the plumbing. UserOperations depend on bundlers, and bundlers are concentrated: a small handful of operators, led by Pimlico and Alchemy, handle most of the volume. That concentration is the centralization complaint that motivates the whole push to make account abstraction native, so that a normal Ethereum node can process a smart-account transaction directly with no privileged middlemen in the path.

The route was pragmatic by necessity. Developers had been circling account abstraction since roughly 2016, and earlier attempts kept running into the same wall: proposals like EIP-2938 needed changes to Ethereum’s consensus rules, which are the hardest kind to ship, while EIP-3074 introduced powerful new opcodes that were eventually set aside. ERC-4337 sidestepped all of that by living entirely at the application layer, which is exactly why it could ship in 2023 without asking anyone to fork the chain. The trade-off it accepted, a separate mempool and a class of privileged operators, is the same trade-off the ecosystem is now trying to undo.

EIP-7702: the upgrade that actually shipped

EIP-7702 was the pragmatic move. Shipped in the Pectra hard fork on 7 May 2025, it added a new transaction type (0x04) that lets an ordinary EOA point at a smart-contract implementation and temporarily behave like a smart account, while keeping its address, its history, and its balance (EIP-7702, Pectra roadmap). No migration, no new address, no waiting for native account abstraction. A short delegation designator (the bytes 0xef0100 followed by a 20-byte contract address) is written into the account’s code slot; delegating to the zero address wipes it clean again. That in-place quality is why wallets adopted it fast. MetaMask made it the primary upgrade path for its Smart Accounts, Ambire was first out of the gate, and Safe, Rabby, Trust Wallet and the exchange wallets of OKX and others followed.

The adoption numbers look enormous and need a caveat. BundleBear shows cumulative authorizations near 246 million and more than 103 million set-code transactions, but only about 57.6 million delegations are live at any moment (BundleBear). The headline authorization figure is heavily inflated by automated sweeper bytecode, which reuses the same delegation across tens of thousands of already-compromised addresses. The honest read is tens of millions of genuinely delegated accounts, not a quarter of a billion distinct users. Even discounted, that is the fastest any account-model change has ever spread on Ethereum.

What smart accounts unlock

Strip away the standards alphabet soup and the point is a handful of concrete features. Batching lets you approve a token and deposit it into a lending market in a single confirmation instead of two, the kind of multi-step flow explained in our guide to borrowing against your crypto. Gas sponsorship lets an app pay your fees, or lets you pay them in USDC. Social recovery lets chosen guardians or a second device restore access if your primary key is lost, without any seed phrase at all. Spend permissions let you pre-authorize an app to pull a fixed amount on a schedule, the on-chain version of a subscription.

Session keys are the feature that most changes how crypto feels. A session key is a scoped, time-limited signer that can perform a narrow set of actions without a fresh prompt each time. In games, a session key lets a title sign dozens of in-game moves without a popup on every action, though each on-chain move can still carry consequences at filing time, as we cover in our look at NFT and gaming taxes. The same primitive now underpins AI agents that hold a bounded spending allowance and act on a user’s behalf, a frontier that security researchers are watching closely, as detailed in our interview with Trail of Bits on agent security. The power and the danger are the same mechanism: an account that can act without asking is convenient for you and equally convenient for an attacker who captures the key.

Who pays for gasless: paymaster economics

Gasless is the headline feature, and someone always picks up the bill. Paymasters come in two broad flavors. In verifying mode the sponsor, usually the app, covers the fee outright as a customer-acquisition cost, the way a game might eat the gas to onboard a new player. In ERC-20 mode the user still pays, just in a token like USDC rather than in ETH, and the paymaster swaps it behind the scenes. Circle’s Paymaster is the clearest example: it lets an account pay gas in USDC on chains including Arbitrum and Base, and it charges roughly a 10% surcharge over the raw gas cost for the service (Circle).

The economics matter because sponsorship is a lever, not a gift. Apps meter it with budgets and rate limits to avoid griefing, where a bad actor drains a sponsorship pool by spamming worthless transactions. Providers such as Pimlico, Alchemy and Biconomy sell paymaster and bundler infrastructure on tiered pricing, and on layer-2 networks the underlying gas for a simple transfer is often a fraction of a cent, which is what makes free-to-the-user flows viable in the first place. For the user, the takeaway is simple: gasless does not mean free, it means the cost moved somewhere less visible.

Passkeys and the seed-phrase problem

The seed phrase is the single biggest barrier to mainstream self-custody, and smart accounts are the mechanism that lets it disappear. The bridge is a standard called RIP-7212, a precompile that lets Ethereum and its layer-2s verify signatures over the P-256 curve cheaply, cutting the cost from roughly 300,000 gas to around 3,450 gas (Alchemy). P-256 happens to be the curve behind Apple’s Secure Enclave, Android’s Keystore, and the WebAuthn passkeys already built into every modern phone and browser. Once an account can validate a P-256 signature, your fingerprint or face can authorize a transaction directly, with the key material sealed in hardware that never leaves the device.

The installed base is already there. The FIDO Alliance reported roughly 5 billion passkeys in use worldwide by its 2026 World Passkey Day, with a large majority of surveyed users having enabled at least one (FIDO Alliance). That is why consumer wallets from Coinbase, Base and others now onboard users with an email or a passkey and no twelve-word backup in sight. The trade-off is that passkey recovery leans on cloud sync from Apple or Google, which reintroduces a custodial dependency at the edges, so most serious wallets pair passkeys with a separate recovery path rather than betting everything on the phone.

The players in 2026

The account-abstraction market has settled into a few clear camps: multisig infrastructure for organizations, consumer smart wallets for individuals, and wallet upgrades that bolt smart-account features onto the incumbent apps people already use. The table below maps the main products and the path each one takes.

Safe remains the institutional heavyweight. Its smart accounts processed nearly 130 million transactions in the second quarter of 2026 and held $27.24 billion in self-custodied assets, including $6.48 billion in stablecoins, with total accounts up about 20% year over year to 63.4 million (The Block, Safe Ecosystem Foundation). Lukas Schor, a co-founder of Safe, noted that the activity “held up even as crypto prices fell sharply in June,” with active accounts still rising through the quarter. That resilience under a falling market is the strongest evidence yet that smart accounts are infrastructure now, not a demo.

Under the hood, most of these products now share a modular design so that features like recovery, session keys and spending policies can be added as plug-in modules rather than baked into one monolithic contract. Two competing standards define how those modules attach. ERC-7579 is the minimal, widely adopted approach used by Safe, ZeroDev and Biconomy, while ERC-6900, championed by Alchemy, is a heavier and more prescriptive framework (EIP-7579). For a user, the practical consequence is portability: a module written for one 7579 wallet can, in principle, work in another, which is the closest the market has come to a shared smart-account plugin ecosystem.

The native endgame, and why it keeps slipping

The whole point of ERC-4337 and EIP-7702 was to buy time until Ethereum could do account abstraction natively, with no side mempool and no external bundlers, so that a plain node could validate a smart-account transaction the same way it validates any other. In 2026 that finish line moved again. Base had planned to ship its native standard, EIP-8130, in its Cobalt upgrade this September. On 8 September a merged pull request, number 4924, re-gated the feature from Cobalt to a new fork called Zenith, and Zenith is set to None across mainnet, Sepolia and every canonical configuration, meaning it is devnet-only with no activation date (base/base PR #4924). Native account abstraction on the largest layer-2 slipped from weeks away to unscheduled.

Worse for anyone hoping for a single standard, Base and Ethereum L1 developers stopped trying to converge. There are now two competing native designs (CryptoBriefing). EIP-8130 registers who can act for an account in an on-chain configuration and validates every transaction against it, with no bundlers or relays, and Base reports a USDC transfer dropping from 125,000 to 46,000 gas, a 63% cut, with transaction size down more than 83% (Base engineering). EIP-8141, the Frame Transactions proposal aimed at Ethereum’s future Hegota upgrade, takes the maximally flexible route: arbitrary validation logic, post-quantum readiness, native gas sponsorship and trustless token payments. Vitalik Buterin has framed a native omnibus as achievable “within a year,” though he has not committed to a fork date (Cointelegraph).

The two designs reflect different priorities, not just different code. Ethereum’s EIP-8141 leans toward censorship resistance, value capture at the base layer, privacy and post-quantum readiness, favoring maximum flexibility even at the cost of complexity. Base’s EIP-8130 optimizes for scalability, gas efficiency and quick adoption across the OP Stack, accepting a narrower, fixed set of options in exchange for a simpler, cheaper design (CryptoBriefing). One camp is building for the endgame of a self-sovereign account that survives even if the surrounding infrastructure fails; the other is shipping something wallets can use next year.

The split is now a genuine debate rather than a scheduling hiccup. Pedro Gomes, founder of WalletConnect (now Reown), wrote that after months working on EIP-8141 he is “convinced EIP-8130 is the better path for native account abstraction,” calling it “simpler, more portable, and focused on what wallets actually need” (@pedrouid). Derek Chiang, founder of the smart-account infrastructure firm ZeroDev, took the other side of the cost, telling CryptoBriefing that “a unified standard would have created a more consistent multichain smart account experience,” while allowing that parallel innovation could still produce something better. The three leading native proposals stack up like this.

Glamsterdam will not be the fix

If native account abstraction is not coming from Base’s next fork, the natural question is whether Ethereum’s own next upgrade carries it. It does not. The Ethereum Foundation launched an early public testnet named Plataberget on 17 August and forked it to the Glamsterdam ruleset on 20 August, giving developers months to break things before the changes reach Sepolia and mainnet (Ethereum Foundation). Glamsterdam is a serious upgrade, but its headliners are enshrined proposer-builder separation (EIP-7732) and block-level access lists (EIP-7928), not account abstraction. Native AA is not on the marquee.

Glamsterdam does bring one change that every wallet team has to reckon with anyway: a coordinated gas repricing toward a roughly 200 million gas floor, which removes the hard-capped maximum gas limit. As the Ethereum Foundation and ecosystem builders have warned, any wallet, indexer or gas estimator that assumes a fixed cap will break (Etherspot roundup). In other words, the next big Ethereum upgrade does not deliver the account-abstraction endgame, but it does hand wallet builders a pile of compatibility work. The gap between where smart accounts are and where they were promised to be is not closing on the schedule anyone drew in 2023.

The fragmentation debate

Underneath the standards fight is a deeper worry: native account abstraction might standardize the account and still leave the user experience splintered. When AA builders gathered to debate the point this month, the sharpest concern was session-key interoperability. If each wallet keeps whitelisting only its own implementation, the same fragmentation that plagues today’s smart accounts simply reappears one layer down, and a permission you granted in one wallet means nothing in another. Standardizing the transaction format does not automatically standardize what apps and wallets agree to honor.

Vitalik Buterin’s contribution, as reported from the discussion, was that each user should keep exactly one master recovery root for their entire on-chain life across Ethereum and its layer-2s, a design he argues enables clean, layered security. Others pushed back that a single root becomes a single point of failure (Etherspot roundup). A blunter observation surfaced too: wallets are effectively loss leaders that earn through in-app swaps and order flow, which gives them little commercial reason to let apps abstract them away. Bad onboarding, in that reading, is not always an accident; sometimes it is a business model. Native AA can fix the protocol, but it cannot by itself fix incentives.

The security ledger: the protocol works, the keys do not

Account abstraction has a strong safety record at the contract level and a troubling one at the human level. The smart-account contracts that matter are heavily audited and the standards have held up in production. The losses come from the signing screen. The largest crypto theft on record, the roughly $1.5 billion Bybit hack of February 2025, was not a flaw in Safe’s contracts; attackers compromised a developer machine, served malicious code through the signing interface, and tricked signers into approving a transaction that swapped the account’s implementation while the screen showed a routine transfer (The Block). It is the defining example of blind signing, where a signer approves a payload they cannot read, and it echoes the pattern in 2026’s other biggest key heists.

EIP-7702’s own dark side follows the same logic. Analysis by Wintermute found that more than 97% of early 7702 delegations reused identical automated sweeper bytecode, nicknamed CrimeEnjoyor, spread across tens of thousands of already-compromised addresses; the operation was not even profitable because the wallets were empty (CoinDesk). A peer-reviewed study presented at the 2026 USENIX Security Symposium examined 3.66 million 7702 authorizations across seven chains and found that more than 63% were tied to malicious contracts, with 924 distinct malicious contracts confirmed and about $2.36 million in realized losses (USENIX). The nuance is that those figures count transactions, not victims, and reflect attackers recycling the same code, not a break in the protocol. The broader trend is even encouraging: total phishing and drainer losses fell 83% in 2025 to about $83.85 million (Cointelegraph). The lesson is consistent across every incident. The protocol works; private keys and the humans holding them are the weak point, which is why a hardware wallet, or signer, that shows you what you are actually approving still matters, as we detail in our 2026 signer reviews.

What it means for exchanges, custody, and the SEC

Smart accounts complicate life for exchanges and custodians in a specific, technical way. Because an EIP-7702 delegation writes a recognizable 23-byte designator into an address’s code slot, deposit and withdrawal systems now screen for it. A deposit address that suddenly carries delegated code could be under a sweeper’s control, and consolidating funds from many user deposit addresses becomes riskier when any of them might auto-forward. The operational answer at institutional desks has been tighter screening, multi-party computation for signing, and allow-lists of trusted delegate contracts, so that a compromised delegation cannot quietly redirect a withdrawal.

For self-custody users in the United States, the regulatory picture got clearer in 2026. Securities and Exchange Commission staff indicated that software allowing users to transact from self-hosted wallets is not, by itself, acting as a broker (CoinDesk). That distinction matters for account abstraction: a smart-account wallet that never takes custody of your keys sits outside the broker framework, while a custodial exchange offering the same features does not. The other side of self-custody remains unchanged and unforgiving: there are no chargebacks. A smart account can add spending limits, recovery guardians and simulation warnings, but if you sign the wrong thing, no regulator and no support desk can claw it back. Account abstraction makes the safeguards better; it does not make the network reversible.

The bottom line

Account abstraction in 2026 is a story of a working present and a delayed future. The interim stack, ERC-4337 for full-featured smart accounts and EIP-7702 for upgrading the address you already hold, is live, widely adopted, and good enough that most users benefit from it without ever learning the acronyms. The native version that was meant to retire that stack keeps slipping: Base moved EIP-8130 onto an unscheduled fork, the industry has split into two competing standards, and Ethereum’s next upgrade does not carry account abstraction at all. For users, the guidance is simple. The features are real and worth using, the security burden still lands on what you sign, and the finish line remains, as it has for years, roughly one upgrade away.

Frequently Asked Questions

What is account abstraction in simple terms?

It means your crypto account is controlled by programmable code instead of a single private key. That code can require several signers, approve transactions with a passkey or fingerprint, set spending limits, bundle multiple actions into one, let someone else pay the gas, or restore access through a recovery path, none of which a plain key-based account can do on its own.

Is account abstraction the same as EIP-7702?

No. Account abstraction is the broad goal of making accounts programmable. EIP-7702 is one route to it: a transaction type shipped in Ethereum’s Pectra upgrade in May 2025 that lets an existing key-based account run smart-account code without changing its address. ERC-4337 is another route, and native account abstraction built into the protocol is the intended endgame.

Do I need to move my funds to use a smart account?

Usually not. With EIP-7702 your existing address gains smart-account features in place, so nothing has to move. ERC-4337 accounts are separate contract addresses, so switching to one means funding a new account. The right answer depends on which wallet you use and which path it takes.

Why is native account abstraction taking so long?

It requires changing Ethereum’s core protocol, which is deliberately slow and contentious. In 2026 the two leading native proposals, Base’s EIP-8130 and Ethereum’s EIP-8141, stopped converging, and Base moved EIP-8130 off its September upgrade onto an unscheduled fork. The interim standards ERC-4337 and EIP-7702 already work well enough that there is little pressure to rush.

Are smart accounts safe to use?

The smart-account contracts themselves are heavily audited and the protocol has held up, but the risk shifted to the signing screen. The largest crypto theft on record, the roughly $1.5 billion Bybit hack in February 2025, tricked signers into approving a malicious transaction they could not read. Most losses come from phishing and blind signing, not from flaws in account abstraction itself.

Marcus Okafor covers wallets, exchanges and account abstraction for HOGE Wire.