The Oct. 29, 2026, Binance theft—estimated by the exchange at about $6.3 billion—was not described as a simple breach of a single computer. Binance’s account, reported by Reuters, depicts a staged intrusion involving employee devices and social engineering. The FBI attributed the theft to North Korea-linked group TraderTraitor, an allegation North Korea has rejected. Meanwhile, blockchain analysts have tracked an extensive laundering operation, but the public reporting leaves important questions open, especially about the precise role of Binance’s internal AI tool.
17
18
32
34
How the attackers allegedly got in
Binance said attackers compromised two software engineers’ computers using different approaches: malicious software concealed in an advertisement and a fake job interview that led an engineer to download a malicious code repository. The attackers then used a third employee’s computer in the theft, according to the company’s account reported by Reuters. These details come from Binance’s investigation; they are not an independently published forensic reconstruction.
32
Binance also said the attackers used an internal artificial-intelligence tool to gain control of its wallets. But early reporting cautioned that the tool’s exact role was not yet clear. The available reporting does not establish that an autonomous AI system planned or carried out the theft.
19
35 That distinction matters: a compromised system that includes an AI tool is not, by itself, evidence that AI independently made the decisions behind an attack.
What the FBI has said about attribution
On Oct. 31, the FBI attributed the malicious activity to TraderTraitor, a North Korean group, and said investigators observed the actors converting stolen assets primarily into ether. That is the U.S. government’s attribution; North Korea has denied responsibility.
17
18
The reporting also connects TraderTraitor to the 2025 Bybit theft: TRM Labs described it as the same group blamed for that $1.5 billion incident. This is an intelligence firm’s attribution, not proof presented in the available sources that the two cases used identical methods.
8
Tracking the laundering—and what the numbers mean
Elliptic reported that the stolen funds moved through more than 20,000 wallets across 16 blockchains and at least 65,000 transactions. Its figures describe activity traced by the analytics firm; they do not establish that every movement was completed or that the funds were cashed out.
40
42
Elliptic’s Nov. 6 findings put about $3.11 billion in addresses linked to the theft, mostly in ether and bitcoin, and said those assets were very likely still held by North Korea. The firm separately reported about $2.85 billion frozen or seized—roughly 44% of its estimate of the total theft. These categories are different: funds associated with an address, funds frozen, funds seized and funds returned to customers are not interchangeable.
13
43
Elliptic also reported that about $124 million in stolen assets had been converted into bitcoin, with much of the newly converted amount coming through swaps involving THORChain. That is a reported stage in the tracing effort, not evidence that the converted funds had been successfully cashed out.
7
38
Binance’s response is separate from asset recovery
Binance said it had set aside around $6.34 billion of company funds to cover losses. It later said it had covered 95% of the stolen tokens and began reopening withdrawals in phases on Nov. 1. Replacing customer assets with company funds is not the same as recovering the crypto taken in the attack.
33
34
After withdrawals resumed, customers had withdrawn a net $4.72 billion as of Nov. 3, Reuters reported. That figure measures customer outflows, not the amount of stolen funds recovered.
18
23
The central lesson in the reporting is that the incident combined an alleged human-targeted intrusion with a complicated digital-asset trail. Investigators and analytics firms have attributed the attack and tracked large flows, while Binance has moved to cover customer losses. But the exact mechanics of the internal AI tool, the final disposition of the traced funds and the legal details of reimbursement are not established by the available reporting.