Binance Says Fake Interviews, Malware and Internal AI Tool Enabled $6.34 Billion Theft
publishers 46articles 47first reported 3 Nov, 15:16 UTCdeveloping since 3 Nov · 9 editions
Binance said attackers spent months infiltrating its systems through malicious software, fake recruitment interviews and compromised internal tools before targeting about $6.34 billion in digital assets on Oct. 29.
According to the exchange’s investigation, the campaign began in June when a Wallet DevOps engineer downloaded a tampered software installer while searching for an AI coding tool. Binance detected the infection, replaced the device and rotated credentials, but said the attackers had already learned about the engineer’s role and colleagues.
In September, another engineer was approached through a legitimate job platform by people posing as recruiters. After video interviews designed to build trust, the employee downloaded a purported coding assignment containing previously undetected malware. The software replaced a legitimate development-tool component, established a backdoor and concealed evidence of the compromise.
Binance said the attackers used the second employee’s computer and an internal AI tool on Oct. 20 to gain control of other engineers’ devices and obtain a container deployment token. On Oct. 29, they accessed a third engineer’s computer, deployed malicious code and initiated unauthorized transactions from five hot wallets.
The FBI has attributed the theft to a North Korean hacking group it calls TraderTraitor. North Korea rejected the finding, calling it a political provocation.
Binance said it had frozen or intercepted about $2.85 billion of the affected assets. The exchange said it would absorb the losses, had replenished 98% of affected tokens and was purchasing the remaining less-liquid assets. It also replaced compromised devices and credentials, rebuilt affected wallet and AI systems, and strengthened employee-device monitoring.
HOW THIS STORY WAS MADE
Written from 8 articles, 6 with the publisher's own text; 8 independent newsrooms once syndicated copies count as one; this version written 30 h after the record first saw the story; the editor kept it as written.
- Publishers
- 46
- Source articles
- 47
- Given to the writer
- 8, 6 with the article's own text
- Independent newsrooms
- 8
- This version written
- 30 h after the record first saw the story
- Second model (editor)
- kept as written
- Publication gate
- passed
- Human review
- none
WHAT THE EDITOR DID
The second model checked the writer's draft against the same sources and published it as written.
Written by a language model from the sources above, then checked by a second model that may only cut, attribute or correct. How it works →