On September 15, 2026, it was revealed that a simple coding error in a helper contract allowed an attacker to drain around $7.8 million from a Gnosis Safe wallet on Ethereum. The victim was the owner of a leveraged position in rsETH — the liquid staking token of the Kelp DAO protocol. The attacker exploited a vulnerability in a custom Safe module connected to Uniswap and gained unauthorized access to the wallet's funds.
According to CertiK, the root cause was a coding error in the helper contract rather than a compromise of the owner's keys. However, the hacker was unable to carry out the full attack: a rapid response from the Kelp DAO team blocked a second attempt to drain another $95 million in rsETH and forced the protocol to suspend deposits and withdrawals until the situation was stabilized.
Key incident figures: $7.8 million drained from a Gnosis Safe through a bug in a custom module; a second attack targeting $95 million in rsETH was blocked; Kelp DAO suspended deposits and withdrawals; the vulnerability was located in a helper contract used for the Uniswap integration, not in Safe's core logic.
Gnosis Safe (now Safe) is a standard multisig wallet solution on Ethereum used by DAOs, treasuries and large holders. Safe's architecture allows custom modules to be connected — external contracts capable of executing transactions on behalf of the wallet within predefined permissions. This flexibility became the attack vector: the affected wallet had a module installed to manage a leveraged rsETH position through Uniswap.
The module's helper contract contained a coding error that allowed an external actor to bypass checks and issue commands on behalf of the wallet. After gaining unauthorized access to move funds, the attacker drained around $7.8 million in assets. The leveraged position amplified the damage: the wallet held rsETH as collateral alongside borrowed assets, so manipulating the position through the module resulted in losses exceeding the wallet's nominal free balance.
Kelp DAO, the issuer of rsETH, responded to the incident almost immediately. The team confirmed that its rapid response blocked a second attempt to drain another $95 million in rsETH. As a preventive measure, the protocol suspended deposits and withdrawals, limiting the attacker's room to maneuver and preventing leveraged positions from being unwound at manipulated prices.
The freeze had a dual effect. On the one hand, it locked some of the stolen assets within the protocol's perimeter and made it more difficult to cash them out through liquidity pools. On the other hand, it temporarily restricted legitimate users, which is a standard incident-response trade-off: speed of containment versus service availability. The Kelp DAO case demonstrated that pause mechanisms built into contracts at deployment are now no less important than code audits.
| Incident Stage | Amount | Outcome |
|---|---|---|
| Initial Gnosis Safe drain | $7.8M | Funds withdrawn by the attacker |
| Second attack on rsETH | $95M | Blocked by Kelp DAO |
| Protocol response | — | Deposits and withdrawals suspended |
Modules are an official mechanism for extending Gnosis Safe: they can hold keys, set limits and execute transactions without requiring multiple signatures for every operation. This is convenient for automating the management of DeFi positions, but any error in a module's logic effectively becomes a backdoor into the wallet. Unlike a multisig threshold, a module with broad permissions does not require approval from multiple owners: a single vulnerable function may be enough.
The September 15 incident continues a series of 2026 attacks targeting Safe infrastructure: in May, a SquidRouterModule exploit allowed attackers to steal around $3.2 million from 86 Gnosis Safe wallets in two waves, also through a module vulnerability. The pattern is clear: attackers have shifted their focus away from the core multisig logic, which is heavily audited, toward peripheral custom code that teams write and connect themselves, often without comprehensive audits or edge-case testing.
The first lesson is that custom modules and helper contracts must go through the same audit cycle as a protocol's core code. A single unchecked function can wipe out the security benefits of a multisig protecting millions of dollars. The second lesson is that incident-response speed determines the scale of losses. Kelp DAO limited the damage to $7.8 million instead of a potential $100+ million solely through the rapid pausing of contracts and freezing of funds.
For users, the incident is a reminder that a multisig is not an absolute guarantee of security when third-party modules are connected to it. Before installing external integrations, users should check their audit status and the scope of permissions they receive. For the industry, the September 15 case reinforces the trend toward mandatory monitoring of modular risks and insurance for custom integrations: the cost of an error in just a few lines of code has once again been measured in millions of dollars, while the line between “convenient automation” and a “backdoor into the treasury” has proved thinner than many teams assumed.
“Program testing can be used to show the presence of bugs, but never to show their absence.”
— Edsger Dijkstra, computer scientist, on the cost of unverified code
