Binance lost about $6.3 billion on October 29, 2026, through its own wallet-signing service. The credentials came from a DevOps engineer’s computer, whoever used them was an attacker the FBI ties to North Korea, and every signature was valid. Checking who authenticated is not enough. Security has to know who is acting, right now.
Key takeaways
- The Binance theft ran through valid credentials and Binance’s own signing service, not through a broken lock.
- It follows the pattern of the $1.5 billion Bybit theft in February 2025: a compromised engineer’s machine, then trusted access used by the wrong party.
- Ask whether you can tell, in every session, if a human or an agent is acting, and whether it is the right one.
What happened at Binance?
$6.3 billion. One laptop.
That’s how the Binance breach of October 29, 2026 is shaping up. The investigation points to a compromised DevOps engineer’s machine as the way into Binance’s wallet-signing service, with attackers executing pre-programmed smart contract automation at rapid speed, all with no evidence of a leaked private key.
Binance’s own account, published November 3, 2026, fills in the path. A fake job interview put malware on one engineer’s work computer. The attacker then used an internal AI tool to reach a DevOps engineer, whose computer held a container-deployment token. On October 29, with that engineer’s credentials, the attacker reached Binance’s internal wallet-signing service. TRM Labs reports that most of the money left within about 20 minutes, and the FBI has attributed the attack to North Korea’s TraderTraitor group.
If your reaction is to shake your head, say “what a shame,” and get on with your day without seriously re-evaluating your own security, you are part of the problem.
Why does this kind of breach keep happening?
While the size of this breach hasn’t been seen before, the breach itself has elements we’ve seen repeatedly. A developer or engineer endpoint, hijacked sessions, trusted access reused by someone who wasn’t who the system thought they were. These are known weak vectors. The largest exchange heist of 2025 (Bybit, about $1.5 billion in February 2025) followed almost the same playbook. We’ve seen this movie before. We keep buying tickets.
Much will be made of Binance’s asset recovery and the industry’s response. But that doesn’t make the fact that this type of breach keeps happening any less obscene.
How did the attacker get past authentication?
Here is the uncomfortable truth: the attacker didn’t need to break in. Authentication was bypassed, not beaten. From one engineer’s endpoint they reached privileged, system-level access: enough to make the exchange’s own signing systems approve transactions of their choosing. Every signature was valid. Every request looked legitimate. It was all done inside an identity that had already been verified.
That’s where most security programs still put their weight:
- More steps at the front door: another MFA prompt, another approval, another policy.
- More forensics after the fact: the timeline, the root cause, the post-mortem.
What should security leaders require now?
If we’re looking for a lesson learned from the Binance breach, it should be this: we must add the requirement of knowing what is happening right now. Not who authenticated this morning. Not who the session token says it is. Who is actually behind the keyboard, the API call, the pipeline change, the withdrawal request, in this moment and the next one.
And it’s no longer just humans. AI agents and automated workflows now hold credentials, move data, and take actions on our behalf. An agent operating under borrowed or hijacked authority is just as dangerous as a person doing the same, and it moves faster.
What should every security leader be asking?
- Is this a human or an agent? How do we continuously verify it’s the right human or agent in this session?
- Is there access or activity that shouldn’t be allowed?
If the honest answer is “we’d find out in the incident report,” that is an unacceptable gap.
Knowing which human or agent is interacting with your systems, continuously and in the moment, must be a requirement on par with authentication and Zero Trust.
The honest caveat
“One laptop” is shorthand. Binance says the attacker did not simply use one engineer’s permissions, that the attack was more complex, and that independent experts will review its findings, so details may change. Verifying the person in a session would not have stopped the fake interview or the malware; those belong to hiring checks and endpoint tools. Its place is the moment stolen credentials are used inside a session by someone, or something, they do not belong to.
What to do about it
- Treat engineer and DevOps machines as privileged. Their tokens open deployment and signing systems.
- List the actions that move value. Withdrawals, signing requests, pipeline changes. For each, ask what would tell you, in the moment, that the right person or agent is behind it.
- Count agents and automated workflows as actors. If they hold credentials, the wrong party can use them.
- Verify the person through the whole session. Mimoto Continuous Verification recognizes and responds in real time when a user is impersonated by a person or an agent, a session is taken over, or a credential is shared or compromised.
Questions answered
How did the Binance hack happen?
Binance says a fake job interview put malware on an engineer’s computer, the attacker moved on to a DevOps engineer’s machine, and stolen credentials gave access to its internal wallet-signing service on October 29, 2026. The investigation is ongoing.
How much was stolen from Binance?
About $6.3 billion. TRM Labs estimates $6.36 billion at the time of the theft and reports that about $2.7 billion had been frozen by October 30, 2026.
Who was behind the Binance hack?
The FBI attributed it to North Korea’s TraderTraitor actors on October 31, 2026, and TRM Labs reached the same conclusion with high confidence.
Were Binance’s private keys stolen?
Binance says it found no evidence of a wallet private key leak. The attacker used stolen credentials and access to the signing service instead.
Was the Binance hack like the Bybit hack?
In pattern, yes. The $1.5 billion Bybit theft in February 2025 also began with a compromised developer machine and ended with transactions approved through trusted signing.
Sources
- Unite.AI, Binance Details Fake Job Interviews and AI Tool Compromise in Hack, November 2026.
- BleepingComputer, Binance links $6.3 billion crypto theft to compromised DevOps laptop, October 2026.
- TRM Labs, North Korea Steals USD 6.36 Billion From Binance in the Largest Hack in History, October 2026.
- The Block, Lazarus appears to compromise Safe developer machine in lead up to $1.5 billion Bybit hack, February 2025.



