Binance traces its $6.34B theft to a fake AI coding tool install, a fake job interview and an internal AI tool

Binance has published its initial report on the 29 October theft from its hot wallets. The exchange says “approximately USD 6.34 billion in digital assets were impacted” and that about USD 2.85 billion has been frozen or intercepted so far. It says it will absorb the loss, not its users, and that 98% of the affected token assets had been replenished as of 31 October.

Binance calls it “a sophisticated, multi-stage campaign” and says it “was not a single isolated breach or a simple phishing incident.” Two of the steps run through AI developer tooling.

The attack path, as Binance tells it

• June 2026: a Wallet DevOps engineer (Engineer A) “downloaded a tampered Homebrew installation in order to install a popular AI coding tool after searching for it on a search engine.” Binance says it caught the anomaly, replaced the device and rotated credentials, but the engineer’s role and colleagues on the wallet infrastructure team were exposed to the attacker.
• September 2026: a second engineer (Engineer B) was approached on a legitimate job platform by someone posing as a Project Director. After an “HR” video interview and a second round with the “CEO” of a fictitious company, he was told to download an “assignment” repository onto his work computer. Its code ran during setup. Binance published the file operations: the malware swapped the Homebrew copy of pyenv for a malicious version.
• 20 October: from Engineer B’s machine, the attacker “used an internal AI tool to gain control over” Engineer A and a third engineer, and found a container deployment permission token on Engineer A’s computer.
• 29 October, 17:15 to 19:56: the attacker logged onto the third engineer’s computer, used Engineer A’s credentials, deployed a container and ran a script that reached the wallet signing service.

The report does not say which AI coding tool was impersonated, or how the internal AI tool gave control over other engineers’ machines. It does not name the attacker. Separately, Binance says its security team detected and prevented 57 suspected North Korea-linked cases in 2026, using techniques including fake job interviews and “malicious software disguised as popular developer tools.”

Coverage differs on the headline figure: The Hacker News writes $6.43 billion and CNBC $6.3 billion; Binance’s own report says about USD 6.34 billion.

What Binance has changed

As of 3 November, Binance says it has rebuilt the affected wallet system in a new environment, rebuilt “the affected internal AI tool and access controls,” replaced affected devices, decommissioned the engineers’ credentials and tokens, and scanned all employee devices. It is engaging independent third-party security firms to review the incident, and says some technical details will stay confidential.

Why this matters

By Binance’s account, every entry point was a developer workflow: installing an AI coding tool from a search result, running an interview take-home on a work laptop, and an internal AI tool that could reach other engineers’ machines. It is the same install-time pattern as the trojanized Claude Code build in Cursor’s marketplace and the fake AI SDK packages on npm, now tied to a record loss. Install AI tools only from their official instructions, run outside code in a throwaway environment, and treat any internal AI tool as an identity whose reach you scope like a deployment token.