About $6.34 billion in digital assets left Binance wallets after a months-long campaign compromised three engineers and reached internal wallet-signing systems. One engineer installed a tampered tool, while another ran a malicious coding assignment delivered through a fake interview. The attackers later used an internal AI tool, though Binance has not explained how. Roughly $2.85 billion was frozen or intercepted.
What Happened
Binance said attackers stole about $6.34 billion in digital assets after a campaign that unfolded over several months and compromised three engineers. The company’s initial incident report described a sequence involving fake recruiters, malicious software downloads, hijacked employee computers, deployment credentials, an internal AI tool and unauthorized access to a wallet-signing service.
The campaign began in June when a Wallet DevOps engineer searched for a popular AI coding tool and downloaded a tampered Homebrew installation. Binance detected that incident, replaced the device and reset associated credentials. The company said the attackers had nevertheless learned about the engineer’s work and colleagues on the wallet infrastructure team.
In September, a second engineer was contacted through a legitimate job platform by someone posing as a project director. The process included interviews with fake company representatives and ended with a coding assignment downloaded to a company-issued computer. The repository contained malware that replaced a legitimate developer-tool component and opened a channel to attacker infrastructure. Binance said its endpoint protection did not detect the malware at the time.
On October 20, the attackers used the infected computer and an internal Binance AI tool to gain control of systems belonging to the first engineer and a third wallet engineer. Binance has not explained how the AI tool was used. The first engineer’s computer contained a token permitting container deployments. On October 29, the attackers accessed the third engineer’s computer, deployed a container using credentials taken from the first engineer and ran a malicious script.
The attackers reached an internal signing service and moved several types of digital assets from hot wallets. They also abused an Ethereum authorization mechanism to give an attacker-controlled contract partial execution rights over a hot wallet. Suspending withdrawals did not stop those on-chain authorizations. Binance reported that roughly $2.85 billion of the stolen assets had been frozen or intercepted.
Why It Matters
The scale of the theft is striking, but the path to the money is the more broadly useful lesson. The attackers did not rely on a single email or one isolated software weakness. They combined information gathered from an early compromise with a convincing recruitment approach, malware on developer machines, an internal tool and credentials that could deploy containers.
Developer workstations became a bridge to critical financial infrastructure. They held access, context and tokens that allowed the attackers to move toward the wallet-signing environment. This pattern applies beyond cryptocurrency. Engineers in cloud, software and infrastructure teams often possess deployment rights and understand the systems that hold an organization’s most important assets.
The response also showed the limits of conventional shutdown actions. Binance said pausing withdrawals alone could not stop unauthorized movements because malicious authorizations already existed on-chain and operated separately from the normal withdrawal infrastructure. The wallet’s multi-party security design also meant there was no single key available to cancel the delegation immediately.
The FBI attributed the theft to the North Korean state-sponsored group TraderTraitor, while North Korea rejected the attribution. Regardless of attribution, the incident demonstrates how financially motivated state activity can combine patient social engineering with technical access to trusted systems.
What to Watch
Binance’s investigation remains ongoing and subject to independent review. A central unresolved issue is how the internal AI tool contributed to the compromise. Until the company provides more detail, it would be unsafe to assume whether the tool itself was vulnerable, misused through stolen access or simply helped the attackers reach other systems.
Further reporting may also clarify how the attackers maintained access across the three engineers, why the deployment token remained available on a workstation and which controls separated employee devices from wallet-signing operations. Those answers will shape the most relevant corrective actions.
Asset recovery and laundering efforts remain active. The supplied reporting said stolen funds moved through about 12,000 wallets on 11 blockchains in under four days. Organizations should distinguish automated fund movement from claims of autonomous AI use; the evidence cited automation but no sign that AI agents made laundering decisions.
What Defenders Should Do
Recruitment controls: tell employees, especially engineers and administrators, that fake interviews may deliver executable assignments. Require coding assessments and unfamiliar repositories to run in isolated environments rather than on ordinary corporate workstations.
Developer tooling: control where installers, package managers and AI coding tools can be obtained. A downloaded tool should not inherit access to production credentials simply because it runs on an engineer’s device. Monitor replacements of developer components and unexpected shell activity.
Credential design: keep deployment tokens short-lived, narrowly scoped and rapidly revocable. Avoid storing powerful reusable credentials on workstations when a brokered or just-in-time process can provide only the access needed for a specific task.
Critical-system separation: isolate wallet signing and comparable high-value services from routine developer activity. Reaching an engineer’s computer should not provide a direct path to deploy code into transaction infrastructure. Require independent approval and strong logging around container deployment and signing-service changes.
AI governance: log access to internal AI tools and define what systems, data and actions they can reach. Because Binance has not explained the tool’s role, defenders should focus on visibility and least privilege rather than assuming a particular AI-specific attack.
Response planning: prepare actions beyond pausing customer-facing transactions. Teams responsible for blockchain systems should understand how to revoke on-chain authorizations, replace affected infrastructure and coordinate with partners that may freeze or intercept assets.
The Larger Lesson
Major incidents often result from a chain of individually manageable weaknesses. An unsafe download exposed organizational context. A fake interview compromised another engineer. Workstation access revealed deployment capability. Trusted systems then allowed the attackers to reach financial operations.
The durable lesson is to protect trust paths, not just endpoints. Developer devices, recruitment workflows, internal AI tools, deployment tokens and signing services formed one connected system even though different teams may have managed them. Security architecture must assume that one layer will fail and prevent that failure from becoming authority over the organization’s most valuable assets.
