Circle blacklisted “Bitget Exploiter 8” at 05:00 UTC Friday. Tether followed hours later through its multisig, and the two issuers locked roughly 99,990 USDC and 218,023 USDT — about $318,000 combined.
That is the number the headlines ran with. It is also 0.08% of a hack now pegged at $387.5 million.
Last week, this newsletter traced the freeze ceiling through the DPRK-linked Xinbi case: issuers can act fast, but only on their own contracts. This week the ceiling moved from theory to test. A larger exchange hack and a bridge lawsuit turned the same structural gap into two separate legal fronts.
Fast, visible, and structurally narrow: the stablecoin freeze captured residual balances while the attacker’s ETH exit stayed beyond any issuer’s reach.
The tension most readers miss is not speed. Circle’s freeze was faster than its response to April’s $285 million Drift hack, when critics including ZachXBT said it let roughly $232 million in stolen USDC move across chains without stepping in. Circle’s position then was that it freezes when legally required.
This time it moved within hours. And it still barely mattered, because the attacker had already swapped.
The wallet held about 170.47 ETH alongside the frozen stablecoins. That ETH stayed untouched. Trackers from MistTrack show other exploiter-linked addresses still hold more than 63,000 ETH.
No issuer can blacklist Ethereum itself. That is not a slow-response problem or a compliance gap to be closed with better tooling. It is the architecture.
USDC and USDT are contracts with built-in freeze functions. Circle can add an address to a blacklist; Tether can confirm a multisig transaction. Neither can do anything to the native asset of a decentralized network.
The attacker understood this before the issuers did. Within minutes of the first unauthorized transfers, a freshly created wallet spent $19.67 million in USDT0 to buy 7,111 ETH in six minutes on Arbitrum, paying roughly 5% above market through UniswapX and 1inch Fusion.
That premium was the tell. Speed mattered more than price. The stablecoins were the freezable part; the ETH was the exit.
By the time Circle acted at 05:00 UTC, the conversion had already happened. The stablecoins left in the wallet were residual — not the point of the operation, just what the attacker had not gotten around to moving.
Bitget CEO Gracy Chen confirmed attackers compromised a backend system in the exchange’s wallet infrastructure, spoofed transaction data, and triggered the authorization process to move funds out. She ruled out a private key compromise. The breach affected hot and warm wallets. Cold wallets remained secure.
The exchange’s User Protection Fund, holding more than $464 million, will cover the loss. Deposits and trading stayed open. Withdrawals were suspended pending a security review.
Chen pointed to IP addresses matching VPN choices by a DPRK group and on-chain signatures consistent with prior North Korean techniques. Chainalysis added stolen-fund labels across its products. Specter linked the stolen XRP through cross-chain bridges to the $24 million AFX hack from July, also attributed to TraderTraitor.
If attribution holds, this is the same playbook the industry has watched since Bybit lost $1.4 billion in February 2025. Convert fast, bridge often, hold the value in assets no compliance system reaches.
The freeze debate keeps recurring because it produces a visible, datable event: Circle acted at 05:00 UTC. Tether followed seven hours later. A number gets attached — $318,000 — and the industry gets a headline that reads like recovery.
It is not recovery. It is a compliance perimeter. The issuers can only freeze what remains inside their own contracts. Everything else — ETH, XRP, AVAX, BNB, XAUT, bridged variants that are not native USDT or USDC — sits outside that perimeter.
That perimeter is narrow for a reason that will not change: stablecoin issuers are not global asset custodians. They are contract administrators. Their freeze power is a function of the token standard they wrote, nothing more.
This is where the KelpDAO lawsuit enters the picture, and why it matters more than another round of freeze commentary.
Evercrest Technologies, the company behind KelpDAO, filed a notice of civil claim in the Supreme Court of British Columbia on Wednesday against LayerZero Labs Ltd., LayerZero Labs Canada Inc., and co-founder Bryan Pellegrino personally.
The claim pleads negligent misrepresentation, negligence, and defamation. It seeks aggravated and punitive damages. It has not been tested in court. No response has been filed.
The allegations are specific. LayerZero told Evercrest in February 2024 that its draft code was “good” and that there was “[n]o problem” using the default configuration, according to the filing. In March 2024, it explicitly directed Evercrest to use a 1-of-1 setup with LayerZero’s own verifier.
That meant a single verifier — LayerZero’s own network — confirmed tokens had been locked on one chain before equivalent tokens were minted on another. One point of failure for a bridge holding significant value.
The filing also says LayerZero warned a separate developer, USDT0, about risks in its default verifier configurations in late 2024 or early 2025, prompting that developer to run its own. Evercrest says it received no comparable warning.
Then the exploit. An attacker placed malware on a LayerZero developer’s computer on March 6, six weeks before any funds moved. The attacker tampered with LayerZero’s nodes so they fed false readings to its verifier. On April 18, the attacker disabled the third-party nodes the verifier also used, so it was told 116,500 rsETH had been locked on Unichain when nothing had.
With one verifier required, the tokens were minted unbacked. The loss: $292 million. KelpDAO paused the bridges within about an hour and blocked a second attempt.
The aftermath is where the liability fight sharpens. LayerZero’s incident statement said the single-verifier setup contradicted a multi-DVN model it had “consistently recommended to all integration partners.” Pellegrino wrote that “[n]obody should be relying on sole DVN.”
Days later, the filing says, LayerZero admitted it had “made a mistake by allowing [its] DVN to act as a 1-of-1 DVN for high-value transactions.”
Pellegrino responded on X: “The claim continues to be meritless. I will meet them in Vancouver and defend myself accordingly.”
The defamation claim turns on those public statements and their timing. Evercrest says LayerZero blamed KelpDAO for a configuration LayerZero itself directed.
The damages include a 2,000 ETH contribution to restore rsETH’s backing, more than $650 million withdrawn since the exploit, and a fall in the KERNEL token that drew regulator and exchange warnings.
The broader DeFi damage was severe. The exploit drained 116,500 rsETH, forced Aave to borrow $300 million to meet withdrawal demand, and erased $20 billion in total value locked across decentralized finance.
This is the pattern. The Bitget freeze attempt and the KelpDAO lawsuit are two ends of the same shift.
Bitget shows that stablecoin blacklisting is fast but structurally narrow. KelpDAO shows where the liability is moving: not to token issuers, but to cross-chain verifiers and bridge infrastructure.
The attacker in the KelpDAO case never needed to touch a stablecoin. The exploit was entirely in the verification layer. A 1-of-1 setup meant one compromised developer machine and one set of tampered nodes was enough to mint unbacked tokens across chains.
No freeze function reaches that. No issuer has jurisdiction over a verifier network. The only venue for allocation is a courtroom.
That is why the freeze-ceiling thread from last issue continues to widen. The Xinbi case showed issuers can act when funds sit in their contracts. Bitget shows how little of a large hack that captures. KelpDAO shows that the largest cross-chain losses will not even involve issuer-controlled assets at the point of failure.
Freeze powers will remain useful for isolating residual stablecoin balances. They will remain fast. They will also remain narrow.
The recovery mechanism for cross-chain losses is becoming litigation. The question is not which issuer blacklists fastest. It is which verifier configuration, which disclosure, and which contractual representation a court will hold responsible when the bridge fails.
LayerZero’s defense will be that KelpDAO chose the setup. KelpDAO’s claim is that LayerZero directed it, endorsed it in writing, and warned others of the same risk without warning KelpDAO.
That is a factual dispute. It will be resolved in Vancouver, not on-chain.
The Bitget attacker, meanwhile, still holds the ETH. No issuer can touch it. No court has jurisdiction over an address. The compliance perimeter held its shape: fast, visible, and small.
The rest of the loss will be covered by Bitget’s protection fund and, eventually, by courts deciding who pays when a verifier fails. The blacklist did its job. It just was not the job most people thought it was doing.