Binance details fake interviews and compromised AI tools behind $6.3bn wallet breach

Binance published preliminary findings on 3 November about the 29 October theft of approximately $6.34 billion in cryptocurrency. According to the exchange, attackers first obtained information about its wallet engineering team in June when a developer downloaded a modified installer while setting up an AI coding tool. The company says it detected that compromise, replaced the device and rotated credentials, but the attackers had already learned about staff and their roles.

In September, another engineer was approached through a legitimate recruitment platform by people impersonating staff at a fictitious company. Following staged interviews, the engineer downloaded what appeared to be a coding assignment. Binance says the repository installed malware that its endpoint protections did not detect at the time.

On 20 October, the attackers allegedly used the compromised workstation and an internal AI tool to reach other wallet engineers. Binance says they found credentials permitting container deployment and subsequently accessed wallet-signing operations. On 29 October, unauthorized transactions drained assets from hot wallets, including XRP and other tokens.

Binance says it froze or intercepted approximately $2.85 billion of the affected assets. It reported that around 98% of impacted token holdings had been replenished by 3 November, while less-liquid tokens were still being sourced. These figures are company statements, not an independent audit of its entire security environment.

Binance says it will absorb the financial impact using its own resources rather than charge affected customers. It has rebuilt affected wallet systems and internal AI tooling, replaced compromised devices, retired relevant credentials and strengthened monitoring. An independent cybersecurity review is planned, but its findings have not yet been published.

The incident highlights risks beyond stolen private keys: malicious developer packages, convincing recruitment approaches and compromised internal permissions can expose exchange infrastructure. Customers should consult official withdrawal-status notices for supported assets and networks rather than assume that a general reopening applies to every transfer route.