On October 29, 2026, the world witnessed the biggest crypto heist on record. The North Korean hacking group TraderTraitor stole about $6.4 billion in crypto from Binance across ten blockchains, most of it in roughly 20 minutes. That is more than four times the $1.5 billion North Korea took from Bybit in February 2025.
TRM Labs attributed the attack to TraderTraitor on the day of the theft, and Chainalysis followed early on October 30. The FBI formally confirmed North Korea’s responsibility on October 31. TraderTraitor’s recent record includes $308 million from DMM Bitcoin in May 2024, the Bybit theft in February 2025, $280 million from Drift in April 2026, and $387.5 million from Bitget in September 2026. According to Chainlysis, North Korea-linked attacks account for about 89% of all crypto hacking losses so far in 2026, up from 15% in 2021. The UN Security Council’s Panel of Experts estimated that hacking provides about half of North Korea’s foreign-currency income, and that stolen funds cover roughly 40% of its weapons programs.
Yet one fact stands out. Within hours, about $2.85 billion, or 44% of the haul, was frozen or seized. When the Bybit heist happened in 2025, the comparable share was roughly 3%. Binance also said it had replenished about 98% of the affected tokens with its own funds. Co-CEO Richard Teng said withdrawal activity had returned almost to normal within 48 hours of incident. This suggests that even as North Korea’s hacks grow bolder and larger, the response from stablecoin issuers and blockchain analytics firms is improving too. Above all, these private actors moved faster than state agencies and law enforcement, and their response worked.
This post outlines what happened, where the stolen money sits now, how it is being laundered, and what the case reveals about both the strengths of networked governance of crypto by nonstate actors and the limits it still has to overcome.
What happened?
The theft took about five months to set up and minutes to execute. According to Binance’s preliminary incident report, the attackers first compromised a wallet engineer’s computer in June through a tampered software installer, then used a fake job interview to plant malware on a second engineer’s work laptop. From there, they turned one of Binance’s own internal AI tools into a path to the computers of the engineers who ran its wallets, where they found a token that let them deploy software. On October 29, they used that access to reach Binance’s wallet signing service and push through unauthorized withdrawals.
Most of the loot was stablecoins and Bitcoin. About 60% of the value was taken from Binance’s wallets on Ethereum, and USDT alone made up roughly $2.5 billion.
List of Stolen Assets
| Asset | Amount stolen | Value at theft (USD) | Blockchain | Can an issuer freeze it? |
| USDT | ~2.49 billion | ~$2.49 billion | Ethereum, Tron | Yes: Tether froze ~1.76 billion |
| BTC | 18,270 | ~$1.47 billion | Bitcoin | No |
| USDC | 823 million | ~$823 million | Ethereum | Yes: Circle froze all of it |
| ETH | ~200,630 | ~$450 million (est.) | Ethereum | No |
| XRP | ~233.5 million | ~$296 million | XRP Ledger | No |
| SOL | ~970,500 | ~$96 million | Solana | No |
| RLUSD | ~88.5 million | ~$88.5 million | Ethereum | Yes: Ripple froze part |
| NEAR | not reported | ~$44 million | NEAR | No |
| AVAX & others | not reported | ~$28 million | Avalanche | Partly |
| TRX | ~46.4 million | ~$16 million (est.) | Tron | No |
| 180+ other tokens | — | ~$550 million (est.) | Ethereum | Varies |
| Total | ~$6.3–6.4 billion | 10–11 blockchains |
Sources: Global Ledger, LearningCrypto, The Crypto Times, Elliptic.
So, where is the money now?
Nine days after the theft, about half the money is frozen. The other half sits mostly untouched in wallets that North Korea very likely still controls, and almost none of it has been turned into cash (Elliptic, as of November 7, 08:00 UTC).
| Status | Value (USD) | Share of $6.43 billion | Type |
| Frozen or seized | ~$2.85 billion | 44% | Almost all USDT and USDC |
| Still in NK-linked addresses | ~3.13 billion | 49% | Mostly BTC and ETH |
| Cashed out | Inconclusive | Inconclusive |
No issuer could freeze the BTC and ETH that make up most of what remains and the bulk of the money is still in the hands of NK-linked addresses. The 18,270 BTC taken from Binance were split across 19 addresses within half an hour of the theft and have stayed there. Nevertheless, the smaller assets – all the stolen SOL and nearly all the TRX – have been laundered, and XRP began moving on November 5, with about $173 million still left. Swaps of these assets have produced about 2,300 BTC (~$175 million ) in new bitcoin addresses. About 88 BTC has gone through coinjoins, which bundle many users’ transactions together so that no one can tell who paid whom. Elliptic expects North Korea to repeat its Bybit pattern: swap the ETH into bitcoin, then cash out through over-the-counter (OTC) brokers and exchanges over several months. That leaves one chokepoint still. To become cash, the funds must pass through centralized services, where they can still be stopped.
Signs of AI-driven laundering?
Chainalysis described “laundering patterns we have not seen before”. In under four days, the stolen funds passed through about 12,000 wallets in more than 40,000 transactions, about one new wallet every 25 seconds. On Tron, the main early route, activity peaked at about 1,300 transactions in one hour, roughly one every three seconds. These activities also spread across services at speed. Within three days, the funds had moved through about 20 automated services, including decentralized exchanges, lending protocols and cross-chain bridges. By November 6, the count had reached more than 20,000 wallets on 16 blockchains and at least 36 services. No human team could plausibly create a wallet every 25 seconds, around the clock, across a dozen blockchains, and the operation rerouted quickly when one path closed. Commentators began calling it the first case of “agentic money laundering”.
However, Elliptic, which has tracked the funds daily, concluded that the laundering is automated but that “nothing we have seen shows an AI agent making decisions.” Its evidence is the mistakes. The trail is full of failed swaps and repeated transactions, the marks of pre-written scripts supervised by people. A script runs fixed steps and repeats or stalls when one fails. What would point to an agent is behavior no one wrote in advance, such as adapting to new conditions or pursuing goals of its own, and investigation have found none. Other analytics firms reached the same view: automated algorithms, not AI agents. Critics argue the AI-laundering story points at the wrong place, because the one confirmed use of AI in this case came during the break-in, not during the laundering. Elliptic also says the evidence cannot show whether people or AI systems chose which services to use. “No evidence of AI” is therefore not proof that no AI was involved.
For governance, though, what matters more than the label is the security controls on the defenders’ side. Monitoring is what allowed the providers to act so quickly and decisively. Binance’s own system raised an internal alert about the large transfers, and Elliptic says its automated alerts have already led to freezes of stablecoins caught in transit. The response, in other words, combined automated systems with human judgement. Whether scripts or agents run the laundering, those controls will have to keep pace as it gets faster.
What the response reveals about private-led governance
The Binance response shows what private-led governance does well. It is fast, and in this case it worked. It also shows where that power ends, and why depending on it should worry us.
Tether froze USDT on Ethereum just 37 minutes after it reached the attacker, a speed no state could match, and Circle froze all 823 million USDC before any of it could be swapped. Ripple and World Liberty Financial froze their own stablecoins, and Arbitrum’s Security Council froze 8,426 ETH and 74.6 million ARB. When some funds reached the exchange Gate, it froze the linked accounts. Analytics firms named the NK attacker within hours, two days before the FBI did. No state acted on that timescale, and the result was 44% frozen instead of 3% in 2025. This power, however, stops where issuers stop, because freezing works only on assets that have an issuer. On Tron, the attacker swapped 234 million USDT for TRX before Tether’s freeze landed, and about $786 million in stablecoins got away in total. BTC, ETH and XRP have no issuer at all. That is why roughly $3.1 billion remains out of reach.
Because private-led governance is voluntary, no one is obliged to act, and some services decline. THORChain kept processing the stolen funds after the FBI asked operators to block them, moving more than $46 million in 72 hours and earning over $200,000 in fees, according to the Wall Street Journal. Its developers say the code gives them no way to block individual transactions. Critics say that claim hides a centralized, profitable operation, and investors in a THORChain-based lending protocol have sued its developers, questioning its claim to be decentralized.
Implication
The hopeful lesson is that as automated hacking grows more sophisticated, the response capacity of crypto issuers and analytics firms is advancing too. Tether and Circle decided within minutes, on their own authority and without any formal process, which addresses to freeze. It worked here because the theft was obvious and the attribution held up. Just as striking, the response went beyond freezing funds to making victims whole. Binance replaced about 98% of the affected tokens with its own money within five days. It is hard to imagine a government-led process achieving either result so quickly.
But these decisions rest on company policy, and the attribution they lean on comes from analytics firms whose methods are proprietary. The same power, used on a weaker case, could freeze the wrong person’s funds with little recourse. The response was also highly concentrated, since most of the 44% came from just two companies. A financial system marketed as decentralized turned out to depend on a few centralized issuers as its safety net.
Most notable of all, laundering now runs at machine speed, and freezes are starting to follow, which sets up an automation race. Elliptic says its automated alerts have already led to freezes of stablecoins caught in transit. Faster freezing will narrow the speed gap, but every step toward automatic freezing also raises the cost of a wrong match and makes human review harder.