$71 Million in Frozen ETH, Two Sets of Victims

What Happened

On Sunday, May 3, 2026, the law firm Gerstein Harrow LLP appeared on Arbitrum DAO governance forums representing two clients: Han Kim and Yong Seok Kim, families holding federal court terrorism judgments against North Korea entered in 2015. The firm filed a New York restraining notice attempting to seize the 30,766 ETH (approximately $71.

1 million at current prices) that Arbitrum's Security Council froze on April 18 after the Kelp DAO exploit. The notice arrives just before Arbitrum DAO was scheduled to vote on the 'DeFi United' initiative, a proposal announced in late April to use the frozen ETH to compensate the Kelp depositors whose $292 million in rsETH was drained on April 18.

The legal foundation: when the Kelp DAO exploit hit, LayerZero's post-mortem attributed the attack to a subgroup of North Korea's Lazarus Group. TRM Labs corroborated the attribution. The Foreign Sovereign Immunities Act includes a terrorism exception (28 U.S.C. § 1605A) that allows victims of state-sponsored terrorism with valid federal court judgments to seize assets owned by the offending state, regardless of where those assets are held, provided they pass through US jurisdiction at any point.

Gerstein Harrow's argument: the frozen ETH is DPRK property because Lazarus is a DPRK state actor. The Kims' 2015 judgments predate Bitcoin's mainstream existence and have been searching for satisfiable assets for over a decade. The Arbitrum Security Council's frozen pool is the first large DPRK-attributable asset cache to surface in a venue with sufficient US-jurisdictional contact for the FSIA exception to plausibly apply.

The Arbitrum Security Council can freeze attacker funds. A New York court can claim them. The Kelp depositors are not parties to either action. Crypto is not just a tool of evasion anymore. It is a field of restitution.

The Legal Lineage Goes Back to Kim v. DPRK

The Kims' theory does not arrive in a vacuum. It descends from a specific line of cases that the D.C. Circuit shaped in Kim v. DPRK in 2014. In that case, the family of Reverend Kim Dong-shik alleged that DPRK officials had abducted, tortured, and killed him after he led missionary work helping North Korean refugees.

The court ruled the family could proceed against North Korea under the FSIA terrorism exception. That ruling, and the line of authority it established, was part of a broader Congress-driven effort to give victims of state-sponsored terrorism a path to recover against regimes that would never voluntarily appear in US court.

The legal theory worked. Default judgments followed. The problem that has dogged terrorism plaintiffs for the decade since is the same problem the Kims face now: getting paid.

North Korea is not a normal judgment debtor. It holds few visible US assets, operates through front companies, and routes money through sanctions evasion, third-country intermediaries, and increasingly through digital assets. Congress created the Victims of State Sponsored Terrorism Fund in 2015 specifically to address this collection gap, financing awards through penalties, settlements, and forfeited property tied to sanctions enforcement.

The fund is a partial answer. It does not eliminate the deeper question of whether plaintiffs can reach assets that pass through decentralized infrastructure but are traceable to North Korean conduct. The Kelp DAO frozen ETH is the first large pool to test that question with on-chain forensic attribution clean enough to make the legal theory plausible to a federal court.

Old Terror Law Meets New Financial Infrastructure

The right framing for this case is not 'first crypto seizure for terrorism judgments' or 'precedent-setting on-chain attribution.' Those framings undersell what is happening. This is a collision between two legal worlds that usually do not meet. The old world says state sponsors of terrorism should not hide behind sovereign immunity, and that victims with valid federal judgments deserve a path to satisfaction wherever the regime's assets can be found.

The new world says value now travels through smart contracts, bridges, and protocol treasuries where ownership can be fragmented across multiple jurisdictions and contested by parties who never appear in any single courtroom. The Kelp DAO case is the first time these two worlds collide on a frozen pool of ETH that both worlds have plausible claims to.

Three categories of crypto seizure have happened before. US Marshals seizures of forfeited Silk Road and similar criminal-case Bitcoin (the basis of the current Strategic Bitcoin Reserve). OFAC sanctions blocking specific addresses tied to designated entities (Tornado Cash, certain Lazarus wallets).

Bankruptcy clawbacks against US-domiciled exchanges. None of those frameworks fit the Kelp situation. The Arbitrum Security Council is not a US entity. The Kelp DAO smart contracts are not US-jurisdictional. The frozen ETH was never in US custody. The connection point is that Arbitrum DAO governance includes US-resident participants whose votes can be enjoined by US courts, and the funds will eventually move through entities (exchanges, custodians, withdrawal pathways) that do touch US jurisdiction.

That contact is enough for the FSIA exception to attach, if the federal court agrees with the legal theory.

Crypto as a Field of Restitution, Not Just Evasion

If the Gerstein Harrow theory succeeds, the implications reach beyond the Kim families. The DPRK has been one of the clearest cases of a state that treats digital finance as an extension of national power. Treasury and DOJ enforcement actions through 2025 and 2026 have repeatedly described North Korean actors laundering stolen crypto through international networks to fund weapons programs and other state priorities.

The same infrastructure DPRK uses for sanctions evasion is now being turned against the regime by terrorism-judgment plaintiffs attaching to its proceeds. If the theory holds, crypto stops being only a tool of state evasion. It becomes a field of restitution.

The arithmetic is meaningful. There are multiple federal court terrorism judgments against North Korea totaling more than $2 billion across all plaintiffs. Lazarus Group has stolen approximately $6 billion since 2017 by US Treasury attribution, and 76% of all 2026 crypto theft is now attributed to North Korea per TRM Labs.

The supply of plaintiff claims is large, and the supply of attributable on-chain DPRK pools is growing faster than the legal mechanism for reaching them has matured. The Kelp case is just the first large pool to test the theory. Future Lazarus-attributed exploits where attacker funds get frozen on-chain will surface additional claimants.

The Kim case is where the doctrine starts. It is unlikely to be where it ends.

What This Does to the Kelp Depositors

The Kelp DAO exploit was the largest DeFi exploit of 2026, draining $292 million in rsETH on April 18. Aave was left holding $246 million in bad debt. Cascading withdrawals took $8 billion off Aave's TVL within 24 hours. Approximately 18% of rsETH's circulating supply went to the attacker.

The 30,766 ETH the Arbitrum Security Council froze represents the only meaningful recovery available, because the unfrozen funds moved through THORChain within 48 hours and are functionally unrecoverable. The 'DeFi United' initiative was the proposed mechanism for distributing the frozen pool back to documented depositors who could prove their lost rsETH balances.

The restraining notice freezes that distribution. Even if Arbitrum DAO were to vote unanimously to disburse the funds to Kelp depositors today, the act of disbursing into wallets the Kim families' attorneys could trace would expose Arbitrum participants to US contempt-of-court exposure.

The most defensible posture for Arbitrum is procedural delay: do not vote, do not disburse, let the case work through New York courts. That delay is measured in months at minimum, more likely years. For Kelp depositors, the practical effect is that the only recovery they had any plausible claim to is now frozen in a legal proceeding that does not include them as a party.

They are not adversaries of the Kim families. They are also not legally distinguishable from any other downstream creditor of frozen DPRK assets, which means their claim sorts behind a federal terrorism judgment regardless of how Arbitrum DAO votes.

What This Means for You

If you self-custody Bitcoin in your own keys on the Bitcoin network, this case is structurally irrelevant to your operational posture. None of the assets in this dispute were held by individual self-custody users. The frozen pool is wrapped Ethereum that was originally collateral backing wrapped rsETH on multiple chains.

The dispute exists because the funds were held in a venue (Arbitrum's Security Council multi-sig) that has the technical ability to freeze and the legal exposure to a US court order that compels participation. Your Bitcoin UTXOs have neither property. Your hardware wallet does not get served with restraining notices.

If you hold any DeFi position, wrapped asset, or ETF share, this case is the strongest data point yet for what 'depending on a custodian' actually means in 2026. Three weeks ago you might have argued that Arbitrum's Security Council freezing the Kelp attacker's funds was a feature: bad actors get caught, victims get compensated, the system self-corrects.

The May 3 filing reveals the cost of that feature. Any pool of funds that can be frozen by a governance entity can also be claimed by a court order against that entity. The same mechanism that enables recovery enables seizure by any plaintiff with a stronger claim than the depositors.

For ETF and custodial Bitcoin holders, the analogous concern is: if your custodian's holdings ever became attributable to a sanctioned entity through any auditing or attribution process, the same FSIA-style legal mechanisms could in principle apply. The principle does not require the holdings to actually be tainted.

It requires the holdings to be in a venue subject to US jurisdiction and the plaintiff's attribution to be persuasive enough to a federal court.

The sovereignty thesis turns on standing, which is what the tagline names: the Kelp depositors are not parties to either action, yet their recovery sorts behind a 2015 terrorism judgment because their funds sit in a venue where a court can rank competing claimants. That is the mechanism worth internalizing.

Pooled, governable assets do not just expose you to a court order; they enroll you, without notice or a seat, in a contest between claimants you never chose, decided by priority rules you do not control. Self-custody of native BTC keeps you the only claimant. There is no governance entity to enjoin, no venue where a stranger's older judgment can be ranked ahead of your own holding.

This case is the May 2026 illustration of what being your own and only claimant is worth.

What to Watch

Watch the Arbitrum DAO governance response. The Security Council faces a procedural choice with no clean option, and the path it picks will set precedent for every future L2 governance entity facing a similar order. Gerstein Harrow's filings in New York courts are the next thing to track, to see whether the theory is tested through full litigation or whether Arbitrum settles by some procedural mechanism.

Other terrorism-judgment plaintiffs against North Korea bear watching, because if the Kim case succeeds, additional plaintiffs will surface with claims against future frozen-attacker pools. The next major Lazarus-attributed DeFi exploit is also a signal, because the terrorism-judgment seizure pattern is now part of the post-exploit calculus and will affect how protocols handle attacker-fund freezes going forward.

And watch the OFAC Treasury response, because the executive branch has its own Lazarus-attributed asset claims through sanctions enforcement that may or may not align with private terrorism-judgment plaintiffs claiming the same pools. The collision between executive-branch sanctions enforcement and private terrorism-judgment plaintiffs over the same DPRK-attributable assets is likely the next legal layer this case opens.

Tracked entities this report names, with total coverage across the archive. Follow one into the Actor Network to see every report it appears in.

Other reports that name the same tracked actors — a different thread than “related,” which follows wording. These are linked by who appears in them.