Binance: Fake Job Interviews and an AI Tool Enabled $6.3B Heist

In September, a second engineer was approached by someone posing as a recruiter. The person arranged calls and several rounds of interviews, shared social media pages to build trust, and finally prompted the engineer to download a fake interview assignment that was actually malware. Binance says its endpoint security tools did not catch this one. The malware replaced the pyenv executable with a malicious version, giving attackers backdoor access and a link to an external server, and it removed traces of the compromise.

By October 20, the attackers had used Binance's own internal AI tool to gain control over wallets tied to the first compromised engineer and a third employee, the company says. They then logged into the third employee's computer on October 29 and launched the theft.

Why it could not be stopped quickly

The transactions continued nearly two hours after the first unauthorized transfers. Binance says its hot wallet used Multi-Party Computation, so no single private key existed that could immediately revoke the malicious EIP-7702 authorizations.

CEO Richard Teng writes that no user is bearing the loss, and that about $2.85 billion in stolen assets has been frozen or intercepted. He says 98% of impacted tokens have been replenished, the remaining 2% is being completed, and withdrawals have broadly resumed. Binance adds that it has prevented 57 suspected North Korea-linked cyber threat cases.

Significant questions remain. Warnings about job interviews that ask candidates to run code on machines with access to employer data have circulated for years, and the postmortem offers little detail on the internal AI tool or why it helped the attackers.