The floor is a lie; only the whale.
366 million transactions in three months. 63% malicious. That’s not a speculative attack scenario. That’s the post‑mortem data from the first quarter of EIP‑7702 deployment on Ethereum mainnet. The upgrade was supposed to give EOA (externally owned accounts) smart contract capabilities without migration. It delivered exactly that. What it didn’t deliver was a security model that matches the speed of the attackers. Here’s the on‑chain evidence, the broken assumptions, and the silent exploit that nobody’s talking about yet.
Context: The Upgrade That Changed the Semantic of “Owner”
EIP‑7702, activated as part of the Pectra upgrade on May 7, 2025, allows any EOA to temporarily delegate its code execution to a smart contract. The address stays the same, the private key stays the same, but the logic that runs when the address is called can be replaced via a signed delegation. This is a fundamental shift: the private key is no longer the sole authority over the account’s behavior. The delegated code can perform arbitrary operations, including re‑delegating to a different contract, draining assets, or pretending to be a benign wallet. The design is brilliant for flexibility—programmable EOAs, native account abstraction without the complexity of ERC‑4337. But the trade‑off is a new attack surface that the industry wasn’t ready for.
According to the USENIX 2026 study that analyzed 228 billion historical Ethereum transactions to identify patterns, the first three months post‑upgrade saw 3.66 million delegate transactions. Of those, over 63% were linked to malicious contracts. The total value exposed: $2.36 million in direct losses, with an additional $10.14 million at risk due to pending delegations. The attackers didn’t need zero‑days. They exploited the very design philosophy of EIP‑7702.
Core: The Broken Chain of Trust
Let me walk you through the forensic evidence. I’ve audited smart contracts since 2017, and I’ve seen integer overflows, reentrancy, and flash‑loan attacks. But this is different. EIP‑7702 doesn’t introduce a bug—it introduces a new class of systemic vulnerability that breaks the existing security assumptions of the entire Ethereum ecosystem.
First, the `msg.sender == tx.origin` guard is dead.
For years, developers used this check to prevent phishing attacks: if the caller is the same as the transaction originator, the call is likely from a user’s EOA, not a malicious contract. With EIP‑7702, a delegated EOA can have its tx.origin set to the original EOA, but the msg.sender can be the delegated contract. The check passes, but the code is controlled by the attacker. The study found 2,400+ contracts that rely on this pattern—most of them in DeFi, bridges, and multi‑sig wallets. Those contracts are now vulnerable to delegation‑based phishing. I’ve seen this in my own audits: a wallet that looks legitimate, passes the old checks, but is actually running attacker‑supplied bytecode.
Second, the “re‑delegation” attack vector.
A malicious contract can, once delegated, re‑delegate the EOA to another contract. This creates a chain of delegations that can be used to hide the ultimate destination of funds. The study identified 500+ CREATE2 contracts that were deployed in anticipation of EIP‑7702 activation, with no on‑chain code yet. These are “sleeping” delegation endpoints—ready to be activated once the target wallet signs a delegation. The attacker can wait for the right moment, then trigger the delegation to drain the account. This is a time‑bomb attack. The on‑chain data shows that many of these CREATE2 contracts have never been triggered, but they exist. The risk is not theoretical; it’s a ticking clock.
Third, the “benign re‑binding” illusion.
Wallets like MetaMask and Safe display the delegated code address. If the code is a known benign contract (e.g., a simple approve‑and‑transfer), the wallet shows “safe.” But the attacker can change the code after the user confirms the delegation. The study found that 24% of all malicious delegations used a “benign” first delegation that was later replaced without the user’s knowledge. The user sees a green checkmark; the attacker sees a green light. My own experience with the 2021 NFT floor analysis taught me that visual signals are the easiest to fake. The same principle applies here.
Fourth, the 63% malicious ratio is understated.
The study classifies a transaction as malicious only if the delegated contract is explicitly flagged. But many delegations use contracts that are “gray” – they don’t steal immediately, but they collect data, simulate user behavior, or wait for a larger balance. The real ratio could be closer to 70–75%. The 366 million transactions include legitimate use cases (e.g., gas sponsorship, batch transactions), but the noise is overwhelming.
Contrarian: Correlation ≠ Causation, and the Market Is Wrong
You’d think with a 63% malicious rate, the price of ETH would have collapsed. It didn’t. The market is pricing this as a minor security incident—a few million dollars lost, a few wallet warnings. That’s a mistake. The real impact isn’t the $2.36 million; it’s the broken trust in the account abstraction narrative. The market is correlating the upgrade with adoption (366M transactions) but ignoring the causation: the upgrade enabled a new class of attack that is cheaper and faster than any previous vector.
Let me be contrarian: the long‑term outlook for EIP‑7702 is still positive, but the short‑term security gap is a major liability. The floor is a lie; only the whale. The whale in this case is the narrative that account abstraction is ready for mainstream adoption. It’s not. The on‑chain data shows that the security model is at least 12 months behind the attack models. The attackers are using automated scripts to deploy malicious delegation contracts, monitor for new delegations, and drain funds within seconds. The defenders—wallet developers, security auditors, infrastructure providers—are still building the first set of tools to detect these attacks.
The contrarian takeaway: The current market reaction (no significant price drop) is a lagging indicator. The real correction will come when a major protocol—Uniswap, Aave, or a bridge—gets exploited because their old tx.origin checks were bypassed. That’s when the market will realize that EIP‑7702 isn’t just an upgrade; it’s a new security paradigm that requires a complete rebuild of wallet‑side verification. The $10.14 million at risk today could be $100 million tomorrow.
Takeaway: The Next Signal to Watch
Here’s what I’ll be watching in the next 30 days: the number of new EIP‑7702 delegation transactions that use a “white‑listed” contract. The study found that 242 malicious contracts were identified. If the ecosystem starts using a distributed blacklist (like a Chainlink‑style oracle for delegation contracts), the attack surface shrinks. But if the number of unique malicious contracts continues to grow exponentially—and it will, because the barrier to entry is zero—then we’ll see a cascading failure of trust.
The floor is a lie; only the whale. The whale is the data. The data says: don’t delegate to any contract you haven’t personally audited. Don’t rely on wallet UI signals. And if you’re a developer, treat every msg.sender == tx.origin check as a red flag. Rewrite your contracts. The upgrade is live, the attackers are active, and the only way to win is to verify, not trust.
