On Wednesday, October 29, hackers stole about $6.3 billion in crypto from Binance, the world’s biggest crypto exchange. This was probably the largest theft of any kind in history, and it took roughly 20 minutes, less time than a moto ride from Kanombe to town at 5pm. The hackers didn’t break through Binance’s defences. They went after the staff who had access to the money. It’s like a clever thief in Kigali: instead of climbing the wall, he befriends the umuzamu at the gate. The attack started in June. A Binance engineer searched online for an AI coding tool and downloaded a poisoned version that contained a virus. Binance’s security team caught it and replaced the laptop. But by then the hackers had already learned who worked on the systems that hold the company’s money. In September, they reached out to a second engineer on a genuine job website, posing as a company that was hiring. The engineer had video interviews with the ‘HR manager’, then with the ‘CEO’. Neither the positions nor the company were real. Next came the usual job test: download this project and set it up on your work computer. The engineer did. Binance’s security software never noticed the malware. On October 20, the hackers used that infected laptop to get into Binance’s internal AI assistant. Through it, they reached the system that signs off on withdrawals, and drained one of Binance’s hot wallets in about 20 minutes. A hot wallet is like the cash in the MoMo agent’s drawer, kept close so customers can withdraw quickly. A cold wallet is the money locked away at the bank. Thieves always go for the drawer. The FBI says the attack was carried out by TraderTraitor, a North Korean government hacking team. About $3 billion of the stolen money was in digital dollars issued by two companies, Tether and Circle. They froze it before the hackers could spend it, a bit like MTN blocking a number before the thief cashes out. Binance then covered every customer’s loss from its own savings. So where does AI fit in? Binance’s own report names it twice. First, the hackers used it as bait. The first engineer was looking for an AI coding tool and downloaded a fake one. Second, Binance’s own AI assistant had access to many of the company’s computers. Once the hackers took control of it, they could move from one engineer’s machine to the next. Picture a house help who holds the keys to every room. Win her over, and you own the house. TRM Labs, a firm that tracks stolen crypto, believes the hackers are ‘making ample use of AI.’ Its reason is that their tricks have become noticeably more advanced. Other security experts think AI helps hackers to research targets and fool employees. In 2025, North Korean hackers were behind 64% of all the money stolen in crypto hacks worldwide. That includes $1.5 billion from Bybit, another exchange, in February 2025. According to TRM, the Binance raid alone brought in about as much as North Korea stole in the six years before it.” Rwanda passed its law on virtual assets (Law No. 023/2026) in May, and it already applies. The Capital Market Authority, which regulates crypto under the law, is still writing the detailed rules, so no crypto exchange can get a licence in Rwanda yet. That timing works in our favour. Binance paid $6 billion to learn this lesson. We can learn it for free, before the rules are final. The new law already requires crypto companies to have strong cybersecurity, to get independent security audits and to report cyber-attacks. The detailed rules now being written should go further. They should limit how much customer money can sit in hot wallets. They should require security checks on the laptops of every employee who can reach the withdrawal systems and treat internal AI tools as part of those systems. And they should add something the law currently leaves out, according to PwC: a compensation fund that pays customers back after a hack. Binance survived this because it had a few billion dollars to spare. A Kigali startup won’t have that. Banks and companies building crypto or AI products, your employe’s laptop can now easily become your front door. Does your team know how to detect malicious AI powered tools? Does your internal AI chatbot touch the systems that move money? If one stolen laptop could move customer money, you’re not ready. Binance lost $6 billion to one fake job interview. Let’s write our virtual assets rules with that in mind. The writer is co-founder of RWAiGHT, an AI advisory firm.
