The attacker got the money out, but a bot got it first—all in the same block.
The Summary
- An attacker drained $7.8 million in rsETH from an Ethereum Safe wallet by exploiting an over-permissioned helper module the wallet owner had authorized
- Security firms traced the breach to the helper contract itself, not to any vulnerability in Safe's core wallet infrastructure
- A MEV bot called Yoink front-ran the attacker in the same block, stealing the stolen funds before the hacker could cash out
- The real story: this wasn't a hack of Safe, it was a permissions configuration error that cost $7.8 million
The Signal
Safe wallets are multi-signature smart contracts designed to require multiple approvals before moving funds. They're the closest thing crypto has to institutional-grade custody. But Safe also supports modules—helper contracts that can automate certain actions without requiring all signers to approve every transaction. That's where this fell apart.
The wallet owner authorized a helper contract with far more permissions than it needed. Security researchers confirmed the breach came from this over-permissioned module, not from any flaw in Safe's code. Think of it like giving your house cleaner the code to your gun safe because it was easier than walking them through the security system every week.
"The attacker exploited a coding mistake in authorization, not a vulnerability in the wallet itself."
The attacker identified the weak module, crafted a transaction to drain $7.8 million in rsETH (Rocket Pool's liquid staking token), and broadcast it to the network. But a bot named Yoink intercepted the transaction in the mempool and executed its own version with a higher gas fee, beating the attacker to the payout in the exact same block.
This is maximal extractable value (MEV) in its purest form: bots scanning every pending transaction for profit opportunities, then paying miners to prioritize their version. MEV bots don't care if you're a hacker or a hedge fund. If there's value to extract, they'll extract it faster than you can blink.
Key takeaways from this incident:
- Safe's core wallet remains secure, the breach was in user-configured permissions
- Helper modules are powerful but require the same scrutiny as any smart contract handling funds
- MEV infrastructure is now sophisticated enough to front-run attackers mid-exploit
The Implication
If you're running a multi-sig wallet with modules enabled, audit every permission you've granted. The mental model most people bring from Web2 doesn't work here. In traditional banking, you can reverse a fraudulent wire. In crypto, once the transaction confirms, the money is gone. And now we know that even if you're the attacker, someone faster might take it from you first.
The rise of MEV bots as a check on bad actors is an interesting emergent property of transparent blockchains, but it's not a feature you want to rely on. The correct move is to never grant permissions you don't fully understand. Every module is a potential attack vector. Treat them like loaded guns.