
RafaVisual artistComing December 1
JolinaMarketingComing December 1
How Binance's Internal AI Opened the Vault
A company assistant meant for coders became a path into payment systems holding billions.
In a nutshell
Intruders penetrated Binance by tricking an employee with a fake job interview, then used an internal workplace AI assistant to hop across staff laptops and reach systems authorized to move customer funds. The intrusion led to the theft of $6.34 billion on October 29, 2026, which U.S. authorities attribute to North Korean state hackers. While Binance has rebuilt its internal software and pledged its own funds to cover customer accounts, it has not yet disclosed why an everyday staff tool had the network permissions to connect ordinary employee computers to its transaction signing systems.
Highlights
- Intruders stole approximately $6.34 billion in digital assets after reaching transaction signing services on October 29, 2026.
- The attackers used an internal Binance AI tool on October 20 to move from one compromised computer to two wallet engineers' laptops.
- The U.S. Federal Bureau of Investigation attributed the multi-month intrusion to North Korea's TraderTraitor hacking group.
- Binance reported that $2.85 billion in stolen funds had been frozen or intercepted by authorities across global networks.
- The exchange announced it had replenished 98% of affected token assets using its own corporate reserves by October 31.
From the Editor’s Diary
Workplace AI programs carry the same risks as human staff: if they hold broad network access, any intruder who slips inside can use them to bypass internal security walls.
Who's involved
Binance
The world's largest online exchange for trading and storing customer digital currencies
goal → Block future intrusions, restore missing customer funds and prove its computer systems are safe
TraderTraitor, attributed to North Korea
A state-sponsored computer hacking group blamed by U.S. investigators for the theft
goal → Siphon funds out of the platform and hide the proceeds across global networks
Engineer B
A Binance software developer tricked into installing hidden intruder software during a fake job interview
goal → Protect his company computer while exploring an external career opening
Engineer A
A Binance systems engineer whose computer was penetrated in June and later stored crucial software deployment passes
goal → Maintain core exchange software and safeguard administrative system keys
Engineer C
A Binance developer whose workstation ran the final unauthorized software package
goal → Build secure wallet tools and stop unauthorized users from hijacking his workstation
Richard Teng and Jimmy Su
Binance's co-chief executive and chief security officer leading public disclosures
goal → Uncover how the breach occurred, reassure depositors and submit internal findings to outside reviewers
Federal Bureau of Investigation
The primary federal law enforcement agency of the United States
goal → Identify the culprits, intercept laundered funds and track down the missing money
Independent cybersecurity researchers
Outside technical specialists who study network breaches and automated software vulnerabilities
goal → Discover hidden security flaws and ensure money-moving systems cannot be hijacked through staff tools
In short
Workplace AI tools can quietly punch holes through company defenses. An internal assistant designed to help programmers at Binance, the world's largest digital currency exchange, became an unintended conduit that let outside intruders leap across employee laptops and reach systems controlling customer funds. The intrusion led to the theft of about $6.34 billion on October 29, 2026, an operation the FBI blames on North Korean state hackers.
Financial companies will face growing pressure to strip corporate AI programs of privileged system permissions. That outcome looks likely because security audits and industry experts are already warning that assistants sharing network connections across employee computers create untracked paths into critical infrastructure.
The theft did not happen because an artificial intelligence decided to steal on its own or suffered a clever conversational trick. Instead, intruders treated the tool as a bridge between separate staff computers, grabbing powerful administrative keys along the way. While Binance says it has rebuilt the affected software and separated its networks, it has not yet disclosed how the assistant possessed the power to reach such sensitive systems.
Previously in this story

Binance leaves $6.3 billion defenses unproven
10 November 2026Binance promises deeper network barriers after a record digital theft, but independent proof of its new security remains pending.
FBI Blames North Korea as Binance Breach Leaves Key Question
7 November 2026Federal investigators named a North Korean state group in the six-billion-dollar Binance theft, but how the attackers authorized the transfers remains unexplained.
Binance loses 6.34 billion dollars after fake job interview
4 November 2026A staged hiring test let hackers reach internal transaction systems in the largest crypto heist on record.Ethereum Validators Pocket $29.5 Million From Binance Hack
1 November 2026Attackers' rushed swaps after a $6 billion Binance breach created price gaps that arbitrage bots fought over, sending millions in tips to Ethereum network operators.Coinbase leads crypto stock slide of 12.54%
31 October 2026Shares tied to digital currencies dropped sharply after a major US trading firm reported lower fee income and an overseas rival disclosed a multibillion-dollar theft
How it unfolded
The First Intrusion Is Detected but Leaves Intelligence Behind
The first incident did not directly breach the money-handling systems. Instead, it gave intruders an early look into Binance's technical ranks before security staff wiped the affected machine. While the company contained that initial infection, the intruders walked away with an internal roster of the engineers who maintained the platform's payment architecture. Binance has not shared the exact files taken or published forensic proof directly tying this early probe to the later autumn attack.
A Fake Recruitment Process Creates a Second Foothold
Three months later, the intruders approached another Binance engineer by posing as headhunters hiring for an outside executive role. They held convincing video calls and sent over a practical programming test that carried hidden malicious software. When the engineer ran the assignment, the code took root on his work laptop while bypassing standard security alerts. That quietly gave the hackers a working outpost inside Binance's network.
The AI Tool Becomes a Bridge Between Engineers
On October 20, the intruders pivoted from that single laptop to other staff machines by using Binance's internal AI assistant as a transit route. The tool allowed them to establish control over laptops belonging to two key wallet engineers, where they found an administrative token that authorized broad software updates. Binance has not yet revealed what permissions the assistant held, whether its core coding had security flaws, or why an everyday workplace tool was capable of reaching across separate employee computers.
Stolen Deployment Authority Reaches Wallet Signing
Nine days later, the attackers executed their final move. Working from one engineer's machine with administrative permissions lifted from another, they deployed unauthorized code into Binance's production network. That gave them direct access to the internal service that signs and approves digital currency transfers, bypassing routine customer checks. On the Ethereum network, the intruders also manipulated account delegation rules, allowing their delegated smart-contract permissions to keep draining funds even after Binance temporarily halted regular customer withdrawals.
Investigators Discover a Larger Intrusion Than First Reported
Binance initially pointed to an isolated laptop breach, but investigators quickly determined that the intrusion was far broader. Crucially, the exchange confirmed that its core master keys were not leaked; the attackers had simply gained access to the software authorized to use them. As the FBI pointed to North Korean operatives, outside security specialists warned that standard developer computers should never hold the power to move corporate funds without independent verification.
Binance Discloses the AI Connection and Promises New Boundaries
On November 3, Binance acknowledged the central role of its internal AI tool and outlined changes to its network defenses. Chief security officer Jimmy Su explained how the multi-month campaign converged on the AI assistant to bridge separate machines. The exchange announced that it had rebuilt its wallet systems, scrapped compromised credentials, and ordered an independent security audit. Even so, the company has yet to confirm whether future automated workflows will require separate human approval before touching payment infrastructure.
A contaminated software installation exposes Binance's wallet team
The earliest trace of the intrusion surfaced in June 2026, when an engineer setting up an automated coding assistant downloaded a booby-trapped software package. The company detected the breach, wiped the workstation, and reassigned account keys, but intruders still managed to copy organizational details about the team that managed customer wallets. Those initial clues went unannounced until Binance released an incident review on November 3, days after on-chain tracking specialists spotted abnormal asset outflows on October 29.
Where things stand
Binance's preliminary disclosures remain the main public account of the intrusion, while an independent forensic breakdown has yet to be released. The exchange has pledged its own reserves to cover user balances and says authorities have frozen $2.85 billion, though freezing funds does not mean they have been returned. The company's statement that 98% of missing tokens were replenished on October 31 reflects internal reserve replacement rather than recovered money.
The critical security question remains unanswered: what specific permissions allowed the internal AI tool to reach across employee computers in the first place? Until independent audits confirm that future deployments require separate human approvals and restricted network rights, financial networks remain vulnerable to internal tools acting as bridges across security perimeters.
