EIP-7702 in the Wild: How Smart Accounts Remade Wallets

Fifteen months after Ethereum’s Pectra hard fork went live on 7 May 2025, the change that reached the most wallets was not a fee cut or a throughput bump. It was a single new transaction type, numbered 0x04, that lets an ordinary Ethereum address borrow the powers of a smart contract without giving up its identity. That transaction type is the payload of EIP-7702, and by August 2026 it sits underneath more than 45 million active accounts, according to the analytics dashboard BundleBear.

Most of those users never typed the number. They noticed that their wallet stopped asking them to sign twice to swap a token, that a game let them play without a pop-up on every move, or that a new app paid for their first transaction so they did not need to buy ETH before doing anything. Those conveniences, and a new class of theft that arrived alongside them, are the real story of EIP-7702 in the wallets and exchanges people use every day.

This piece follows the standard from spec into production: the three capabilities that convinced wallet teams to adopt it, the adoption data 15 months on, the drainer economy that weaponized the same feature, and the custody questions it leaves open. For the byte-level mechanics of how an externally owned account becomes a smart account, our companion explainer on how EOAs became smart accounts covers the wiring; here the focus is on what changed for the wallet in your pocket and the exchange holding your coins.

What EIP-7702 Actually Does

Before Pectra, Ethereum split the world into two account types. Externally owned accounts, or EOAs, were controlled by a private key and could do one thing per transaction. Smart contract accounts could batch actions, set spending rules, and sponsor gas, but they were not controlled by a simple key and could not start a transaction on their own. EIP-7702 collapses that divide by letting an EOA point at contract code for as long as the owner wants, while keeping the same address and the same key.

The mechanism is compact. A new transaction type, 0x04 (the specification calls it SET_CODE_TX_TYPE), carries an authorization list. Each entry in that list is a tuple of six fields: chain_id, address, nonce, y_parity, r, and s. When an account signs one of these authorizations, the network writes a short marker into the account’s code slot: the three bytes 0xef0100 followed by the 20-byte address of the contract the account wants to behave like. That 23-byte string is the delegation designator, and from that point on, calls to the EOA run the delegated contract’s logic.

Two details matter for safety, and both come straight from the specification. Setting the delegation target to the zero address clears the marker and turns the account back into a plain EOA, so delegation is reversible. And an authorization signed with chain_id set to 0 is valid on every EVM chain at once, which is convenient for multi-chain setups and dangerous if a user does not realize a single signature now travels. Each authorization is signed over a message prefixed with the magic byte 0x05 and costs 12,500 gas, with an extra 25,000 for a fresh account.

The Alternative That Won

EIP-7702 did not appear from nowhere. For years, Ethereum researchers argued over how to give EOAs smart-contract features. One camp backed EIP-3074, which introduced two new opcodes and a system of invokers. Vitalik Buterin and several co-authors worried that 3074 baked too much special-purpose machinery into the protocol, and in May 2024 they published 7702 as a leaner alternative that added a transaction type instead of new opcodes. The Block reported at the time that the aim was to reconcile competing account-abstraction approaches before the next hard fork.

The pitch was deliberately modest. The EIP’s own text says it “focuses on adding short-term functionality improvements to EOAs which will allow UX improvements to permeate through the entire application stack.” Buterin has framed 7702 as a bridge rather than a destination: forward-compatible with a future in which every account is a smart contract, without committing the protocol to any single design. The plain key-controlled account is meant to fade out gradually, at each user’s own pace, rather than through a forced migration.

Crucially, 7702 does not compete with ERC-4337, the account-abstraction standard whose EntryPoint contract has been live since 2023. It complements it. ERC-4337 built a parallel transaction system (user operations, bundlers, and paymasters) entirely in smart-contract land. EIP-7702 lets the vast base of existing EOAs opt into that world without migrating to a new address, which is why wallet teams treated it as the on-ramp they had been waiting for.

Three Superpowers Wallets Could Not Get Before

Ask a wallet engineer why 7702 mattered and you get the same three answers: batching, gas sponsorship, and session keys. None of them were new ideas; smart-contract wallets had offered versions of each for years. What changed is that they became available to a normal address, the kind a user already had, with funds already in it, without moving anything.

• Batching: many actions in one atomic transaction, so an approval and a swap, or a swap and a stake, either all happen or none do.
• Gas sponsorship: a third party can pay the network fee, so a user with no ETH can still transact.
• Session keys: a temporary, tightly scoped key that signs a limited set of actions for a limited time, so an app stops interrupting every click with a signature prompt.

The next three sections take each in turn, because each reshaped a different corner of the wallet and exchange experience.

Batching and the End of Approve-Then-Swap

The canonical DeFi ritual used to take two transactions. First you approved a token, granting a contract permission to move it. Then you performed the swap, deposit, or trade. Two signatures, two gas payments, and a window in between where the approval sat live on-chain. With 7702, a wallet can bundle both into one atomic action. Approve and swap become a single click; approve, swap, and deposit into a lending market become one as well.

Atomicity is the quiet win. Because the whole bundle succeeds or reverts together, users stop leaving dangling approvals, the standing permissions that drainers have farmed for years. A wallet can grant exactly the allowance a single trade needs and revoke it in the same transaction. That matters most in the composable corners of DeFi; our coverage of modular lending markets and curator risk describes exactly the kind of multi-step deposit flow that 7702 batching now folds into one confirmation.

Exchanges noticed too. A withdrawal that also unwraps a token, or a deposit that routes through an approval, can be presented to the user as one step. For self-custody wallets attached to exchange apps, batching is the difference between a five-tap flow and a one-tap flow, and taps are where users drop off.

Gas Sponsorship and the Paymaster Economy

The second superpower solves the oldest onboarding paradox in crypto: to make your first transaction you need ETH to pay for gas, but to get ETH you usually need to make a transaction. Gas sponsorship breaks the loop. Through a paymaster, an app, exchange, or wallet provider can cover the network fee on a user’s behalf, so a brand-new address with a zero balance can still act.

The economics are simple to describe and harder to sustain. Someone still pays the validator; sponsorship just moves the cost from the user to the app. For a game or an exchange courting new users, absorbing a few cents of gas per onboarding is a marketing expense, cheaper than most ad clicks. Sponsorship can be metered, rate-limited, or tied to a token balance, and it pairs naturally with ERC-4337 paymasters, which already handle the accounting. Alchemy, whose infrastructure powers much of this plumbing, has argued that 7702 is what finally lets mainstream wallets offer gasless transactions to their existing users rather than only to new smart-contract accounts.

A market backdrop helps. With ETH changing hands around $1,915 in early August 2026, per CoinGecko, and base-layer fees well below their bull-market peaks, the cost of sponsoring a transaction is low enough that many apps now treat it as a default rather than a premium feature.

Session Keys: Sign Once, Trade for an Hour

The third capability is the least visible and, for some categories of app, the most transformative. A session key is a temporary signing key with a narrow mandate: trade up to a set amount, touch only a specific contract, and expire after an hour or a day. The user signs once to authorize the session, and for its lifetime the app can act within those limits without prompting for another signature.

For on-chain games, that removes the single worst part of the experience, a wallet pop-up on every move. For trading interfaces, it enables responsive order flow without handing over the master key. Ambire, which bills itself as the first wallet to ship 7702 for everyday users and was a contributor to the standard, has built session keys and account recovery directly on top of delegation. The scoping is the safety feature: a leaked session key can only do what the session allowed, and only until it expires.

Session keys also change how exchanges think about delegated trading. A scoped key that can place orders but not withdraw is a far safer object to hand an automated strategy than an API key with blanket permissions, and it lives on-chain, where its limits are enforced by code rather than by a server.

The Adoption Numbers, 15 Months In

Numbers put the shift in perspective. In the first week after Pectra, The Block counted more than 11,000 EIP-7702 authorizations. Fifteen months later, the aggregate picture from BundleBear looks like this.

The gap between roughly 214.9 million cumulative authorizations and 45.1 million live accounts is not an error. A single account can re-delegate many times, switching implementations or resetting to a plain EOA and back, and each of those counts. What the live figure captures is the number of addresses that, right now, are running delegated code. That is a large share of active Ethereum users, achieved in little over a year, and it happened without asking anyone to move their funds to a new address.

One caveat is worth stating plainly: a meaningful slice of the early authorizations were not human users at all, but automated sweeper contracts, a point the security section returns to. Headline adoption counts and genuine user adoption are not the same number.

Which Wallets and Exchanges Shipped It

Adoption on the user side ran through the wallets. Within weeks of Pectra, several major wallets had 7702 support in testing or behind feature flags, and the rollout has widened since. MetaMask, the largest EOA wallet by active users, built its Smart Accounts Kit (previously the Delegation Toolkit) so existing accounts can opt into delegation and developers can define readable permissions on top.

Exchange-linked wallets were among the earliest heavy users. In the weeks after launch, Dune dashboards showed self-custody wallets tied to OKX and WhiteBIT near the top of the authorization charts, a reminder that exchanges feel UX friction as directly as any app. The distinction that matters for users is between these self-custody wallet products, where you still hold the key, and an exchange’s custodial account, where the exchange does; 7702 operates on the former.

EIP-7702 vs ERC-4337 vs EIP-3074

Because these standards are easy to confuse, here is how they line up. All three aimed at account abstraction; only two shipped, and they now work together.

The practical reading is that 7702 and 4337 are two halves of one system. ERC-4337 built the smart-account infrastructure; EIP-7702 lets a normal address plug into it. Many production setups delegate a 7702 EOA to a 4337-compatible implementation, so the account gains bundler support, paymasters, and validation logic that were designed for smart accounts from the start.

CrimeEnjoyor and the Sweeper Bytecode

Every powerful primitive attracts abuse, and 7702 attracted it fast. Within weeks of Pectra, the trading firm Wintermute noticed something strange in the delegation data: the overwhelming majority of authorizations pointed at the same tiny piece of bytecode, copied across tens of thousands of addresses. Wintermute reverse-engineered it, named it CrimeEnjoyor, and reported that more than 97% of early 7702 delegations ran this one reused sweeper program.

A sweeper does what it sounds like: the moment ETH lands in a compromised address, the delegated code forwards it to the attacker. These are not accounts that were hacked through 7702; they are addresses whose private keys were already stolen, and 7702 simply gave thieves a cheaper, more automated way to drain any funds that arrive. Wintermute’s blunter finding was that the scheme was barely making money, because compromised wallets are usually empty. The firm was careful to separate misuse from a flaw, later stating that EIP-7702 is not flawed, and it is safe to use.

Security researcher Taylor Monahan made the same point from the user’s side. Writing about the sweeper wave, she argued: “It’s not actually a 7702 issue, its the same issue crypto has had since day one: end users struggle to secure their private keys.” The standard, she added, “just unlocks a bunch of cool abilities that make sweeping addresses more cost efficient and less tedious” (as reported by The Block).

When a Single Signature Drains Everything

The more serious threat was not the sweepers but a new phishing pattern that turned batching against users. Because a 7702 transaction can bundle many actions behind one signature, a malicious site can present what looks like a routine approval and hide, inside the same bundle, a delegation to attacker-controlled code plus a series of transfers. One confirmation, and the wallet empties.

The losses were real. In August 2025, the anti-scam service Scam Sniffer documented a victim who lost $1.54 million in a single 7702 phishing transaction, including wrapped staked ETH and wrapped Bitcoin, after signing a batch dressed up as a Uniswap interaction. Two days earlier, another user had lost roughly $1 million to the same crew, and a smaller victim had been hit for about $146,000 weeks before that. Established drainer kits, including Inferno Drainer and Pink Drainer, folded 7702 payloads into their toolchains.

The root cause is blind signing: users approving a transaction they cannot fully read. A batched 7702 payload is denser than a simple transfer, and a hardware wallet that shows only a hash gives the user no chance to spot the trap. This is the same failure mode that has haunted key management for years; our post-mortem of the Coldcard exploit is a reminder that a signing device is only as safe as what it lets you verify. With 7702, the stakes of a single careless signature went up, because one signature can now do far more.

What Auditors and Wallets Check Now

The response has matured quickly, on both the contract side and the wallet side. Auditors reviewing 7702 implementations now work through a specific list of failure modes, and wallet teams have added defenses aimed at the phishing pattern above.

• Storage collisions: because a delegated EOA and its implementation share the account’s storage, a careless upgrade can overwrite critical slots. Auditors push for namespaced storage (the ERC-7201 pattern) to keep layouts from clashing.
• Initialization front-running: 7702 delegation carries no constructor, so any setup step must be protected or an attacker can call it first.
• Cross-chain replay: an authorization signed with chain_id 0 is valid everywhere, so wallets are urged to scope signatures to a single chain unless the user truly wants otherwise.
• Blind signing: wallets increasingly decode and display exactly what a delegation does, warn when an account is about to point at unknown code, and keep allowlists of vetted implementations.

The recurring lesson is sobering: an audit is not a guarantee. Code that passed review has still been exploited through logic no reviewer flagged, a pattern we examined in a look at audited projects that got hacked anyway. For 7702, the defense that matters most is the one closest to the user: a wallet that refuses to let you sign what you cannot see.

The Custody Question Regulators Have Not Answered

Delegation raises a question no regulator has squarely answered: when your address runs someone else’s code, who is in control? For self-custody that is mostly philosophical, because you can revoke the delegation at any time by pointing back at the zero address. For products that sit between users and their keys, it is more than philosophical.

In the United States, the SEC has not issued guidance specific to EIP-7702, and nothing here should be read as suggesting it has. But the standard sharpens existing debates. If an exchange or wallet provider defines the implementation a user’s account delegates to, and can influence what that code does, the clean line between self-custody and custodial control blurs. Session keys, in particular, let a third party act within limits the user set, which is neither full custody nor pure self-custody but something in between.

None of this sits on any regulator’s near-term calendar as a 7702 matter specifically; the deadlines that will actually shape crypto policy this year are elsewhere, as our rundown of the 2026 regulatory countdown lays out. But as smart-account features become the default rather than the exception, expect custody definitions written for a simpler account model to be tested.

What Comes Next: Fusaka and the Endgame

The next Ethereum hard fork, Fusaka, is focused on scaling the data layer and does not change 7702’s semantics; none of its currently scoped changes touch how delegation works. That stability is itself a signal. Fifteen months in, the standard is settled enough that the ecosystem is building on top of it rather than around it.

The trajectory the authors sketched still holds. EIP-7702 was never meant to be the final form of Ethereum accounts; it was meant to carry a huge base of existing users toward a world where every account is a smart account, at their own pace, without a forced migration. The endgame it points at, in which account abstraction is universal and the plain key-controlled EOA fades out, arrives gradually, one delegation at a time. By that measure, the 45 million live smart accounts are less a finish line than a progress bar.

The Bottom Line

EIP-7702 did what infrastructure upgrades rarely manage: it changed the daily experience of using a wallet without asking users to do anything differently, or even to understand what happened. Batching, sponsored gas, and session keys are now table stakes for a competitive wallet, and the exchanges building self-custody products have adopted them because friction is where they lose people.

The same power that made wallets smoother made theft slicker, and the drainer wave is a permanent feature of the terrain now, not a launch-week anomaly. The standard, as its authors and independent researchers agree, is not the problem; the problem is the old one of people signing things they do not understand. The wallets that win the next phase will be the ones that make a 7702 signature as legible as it is powerful. For readers who want the underlying mechanics in full, the companion explainer on how EOAs became smart accounts remains the place to start.

Frequently Asked Questions

What is EIP-7702 in simple terms?

EIP-7702 is an Ethereum standard, live since the Pectra upgrade in May 2025, that lets a normal wallet address temporarily run smart-contract code while keeping the same address and private key. It gives ordinary accounts features that used to require a separate smart-contract wallet: bundling several actions into one transaction, letting someone else pay the gas, and issuing limited-time keys for apps.

Is EIP-7702 safe to use?

The standard itself is widely regarded as safe; the trading firm Wintermute, after studying the first wave of misuse, stated that EIP-7702 is not flawed and is safe to use. The risk is not the protocol but phishing: because one signature can now authorize many actions, a user who signs a malicious batched transaction can lose everything at once. Use a wallet that decodes and displays what a delegation does, and never sign a transaction you cannot read.

Which wallets and exchanges support EIP-7702?

By mid-2026, support spans much of the market. Ambire was among the first, and MetaMask brought delegation to the largest existing user base through its Smart Accounts Kit. Rabby, Trust Wallet, and the smart-account platform Safe have added support, and exchange-linked self-custody wallets such as OKX Wallet and WhiteBIT were early high-volume users. Support applies to self-custody wallets, where you hold the key, not to custodial exchange balances.

Does EIP-7702 replace ERC-4337?

No. The two are complementary. ERC-4337 built a full smart-account system, with user operations, bundlers, and paymasters, that has run since 2023, while EIP-7702 lets existing wallet addresses opt into that system without moving to a new address. Many setups delegate a 7702 account to a 4337-compatible contract, so the account gains bundler and paymaster support.

How do I protect my wallet from EIP-7702 phishing?

Treat every batched signature as high-stakes. Use a wallet that shows the exact contract your account would delegate to and warns on unknown code, avoid signing on sites you reached through links or ads, and be suspicious of any interface that asks you to approve a bundle you did not initiate. If your account is delegated to code you do not trust, you can reset it by delegating to the zero address, which returns it to a plain EOA.

By Marcus Feld, wallets and infrastructure correspondent at HOGE Wire. This article is informational and not financial advice.