Mentioned in This Article
1. About $6.45 Million Leaves a Safe Wallet in One Transaction
A Safe wallet reportedly lost about $6.45 million in cryptocurrency on Monday, November 16, 2026, in an incident that joins a run of recent crypto hacks. Johannes Moormann, co-founder of Zodiac, said a likely compromised key and the wallet's own permission settings allowed the transfer, and that no flaw in Zodiac's smart contracts was involved.
The suspected theft was reported by blockchain security firm PeckShield, which cited the onchain investigator known as Specter. Ethereum records show roughly 4.39 million USDC and 1.72 million syrupUSDC moving out of the wallet in a single transaction on Monday.
USDC is a dollar-pegged stablecoin, and syrupUSDC is a yield-bearing token built on top of it. Both are the kind of dollar-denominated holdings that an organization might keep in a treasury wallet, and the fact that both left in one transaction indicates the transfer was a single authorized action instead of a series of smaller withdrawals.
Who owns the wallet has not been established. Safe and Gnosis Guild were asked for comment and had not responded. The account of what happened so far rests on onchain data, the alerts issued by security researchers, and the review carried out by Zodiac's team.
2. How Did the Safe Wallet Theft Happen?
The transaction went through the wallet's Zodiac Roles Modifier, which checked it against the permissions policy the wallet's operators had configured and approved it. That is the conclusion Moormann said his team reached after reviewing the transaction.
In other words, the system did what it had been set up to do. The modifier did not malfunction and was not bypassed. It evaluated the request against the rules in place and, because the rules permitted the action, let it proceed.
"There is no vulnerability in the module. The contracts worked correctly," Moormann said. He traced the origin of the transaction to a specific participant in the wallet's setup: "The transaction came from the key of a long-standing role member, which was most likely compromised."
Two conditions therefore had to be present together. An attacker needed control of a key belonging to someone who held a role on the wallet, and that role needed to be allowed, under the wallet's policy, to move the funds in question. According to Moormann, both were true. He said the wallet's custom permissions policy permitted the transaction, almost certainly without its operators intending that it should.
3. What the Zodiac Roles Module Does
Safe is a smart contract wallet widely used by teams and organizations that want more than one person to be involved in controlling funds. Rather than a single private key, a Safe is governed by rules written into a contract, and those rules can be extended with add-on modules.
The Zodiac Roles module is one such extension. It lets Safe users assign permissions for specific transactions, so that a given key holder can be authorized to carry out certain defined actions without needing the full set of approvals the wallet would otherwise require.
The appeal of this design is operational. Teams that manage funds actively often need routine actions to be performed without gathering every signer each time. A roles system lets them delegate narrowly defined powers to particular members while keeping broader control behind the wallet's main approval process.
The trade-off is that security then depends on how tightly those delegated powers are drawn. A role scoped to a handful of harmless actions limits the damage a stolen key can do. A role drawn too broadly gives whoever holds that key the same latitude, whether or not they are the intended user. Specter was the first to flag the Zodiac Roles module as part of the suspected attack, and Moormann's explanation locates the problem in the configuration rather than in the module's code.
4. Who Controlled the Keys and Permissions?
Responsibility for the setup, Moormann said, lay entirely with the people running the wallet. The wallet's operators controlled its keys and its permissions, while Zodiac had no access to either and played no part in configuring them.
He also said Zodiac has not been in contact with the affected team. With the wallet's owner unidentified, the review Zodiac carried out was based on the transaction itself and the publicly visible state of the wallet's configuration, not on any exchange with those who lost the funds.
Moormann further addressed the question other users of the module were likely to ask. The incident, he said, does not point to a vulnerability that would affect other wallets using Zodiac Roles. Because the transfer depended on one wallet's particular policy and one member's key, the exposure is specific to that setup rather than shared by every deployment of the module.
That distinction matters for how the event should be understood. A bug in widely used contract code can put every user of that code at risk at once and typically requires a coordinated fix. A compromised key combined with a permissive policy is a failure of operational security at a single organization, and the remedy lies in how keys are stored and how roles are scoped.
5. Where the Funds Went
Tracing by two separate firms shows part of the proceeds heading toward tokens linked to Monero (XMR), a cryptocurrency designed to obscure transaction details. PeckShield said the attacker used about $813,000 to buy Monero-linked tokens on Hyperliquid.
Blockchain analytics firm Bitquery followed the money as well. It traced funds from the Safe wallet through multiple addresses to Hyperliquid, where five accounts bought about $1 million in XMR1 tokens.
The two firms' figures describe the same destination from different vantage points. PeckShield's number refers to what it attributed to the attacker's purchases, while Bitquery's refers to buying by five Hyperliquid accounts that received funds traced from the wallet. Both place the activity on Hyperliquid, and both identify Monero-linked tokens as what was bought. Neither figure accounts for the whole of the roughly $6.45 million that left the wallet.
Moving stolen assets toward privacy-oriented tokens is a common step after a theft, since it complicates efforts to follow the funds further. Stablecoins such as USDC can be traced across public ledgers, and their issuers have the technical ability to freeze balances, which gives a thief an incentive to convert them quickly into assets that are harder to track or restrict.
6. Bitquery Sees Similarities With the Lighter Incident
Bitquery's analysis raised the possibility of a connection to a separate event on the same day. The firm noted that the movement of funds after the Safe wallet transfer resembled the movement of funds following Monday's incident at Lighter.
The resemblances it identified were specific: a transfer of 160 Ether (ETH) and purchases of XMR1 tokens appeared in both sets of activity.
Analysts look for this kind of repetition because the way funds are moved after a theft can act as a signature. Recurring transfer sizes, the same choice of venue and the same assets being purchased can suggest a shared playbook. They can also arise by coincidence, or because different actors copy methods that have worked before, which is why matching patterns alone are treated as a lead and not as proof.
Bitquery was careful about how far that observation goes. It found no transfers linking the two groups of wallets to each other, and it said it could not establish whether the same actor was behind both. The overlap is a pattern in behavior, not evidence of common ownership, and for now the two incidents remain separate cases that happen to share certain features.
7. Why Permission Settings Matter in Crypto Hacks
The episode illustrates a category of loss that involves no broken code at all. When people think of crypto hacks, they often picture an attacker discovering a programming error in a smart contract. In this case, by Moormann's account, every contract performed exactly as written.
What failed instead were the two elements that sit around the code. The first is key management: a credential belonging to a long-standing role member was, in Moormann's words, most likely compromised. The second is policy design: the rules attached to that member's role were loose enough to authorize a transfer of roughly $6.45 million.
Either safeguard alone might have prevented the loss. A stolen key attached to a tightly limited role would have given an attacker little to work with. A broadly empowered role whose key stayed secure would never have been abused. The loss occurred because both weaknesses were present at once.
The description of the key holder as long-standing is also worth noting. Permissions granted early in a wallet's life tend to stay in place unless someone revisits them, and a role that made sense when it was created can outlast the circumstances that justified it. Moormann did not say how long the policy had been configured as it was, only that it almost certainly allowed more than its operators meant it to.
For organizations using delegated permissions on shared wallets, the practical implication is that the configuration deserves the same scrutiny as the contracts themselves. A policy that authorizes more than its authors intended will be enforced just as faithfully as one that is drawn correctly.
8. What Remains Unknown
Several questions are still open. The identity of the wallet's owner has not been determined, and neither Safe nor Gnosis Guild had responded to requests for comment. How the role member's key came to be compromised has not been explained.
Attribution is also unresolved. Bitquery's finding of similar fund flows after the Lighter incident stops short of tying the two events to one actor, and no transfers connect the wallets involved.
Until the affected team comes forward or investigators publish further tracing, the picture rests on what the blockchain shows and on Zodiac's review: a single transaction on November 16, about 4.39 million USDC and 1.72 million syrupUSDC removed, a roles policy that allowed it, and a portion of the proceeds routed through Hyperliquid into Monero-linked tokens.
