This story is currently available in Englisch only — a translation isn't ready yet.
AI Has Made Crypto Theft Cheap: $8.7 Billion Stolen in 2026
According to TRM Labs, $8.7 billion in crypto was stolen across 325 incidents this year; the withdrawal of $6.3 billion from Binance took about 20 minutes.
It took North Korean hackers months to breach Binance, but withdrawing more than $6.3 billion took about 20 minutes. According to the report compiled by Ekonomim from Forbes Türkiye, the operation began long before the October 29 heist. The attackers used compromised software and a fake job interview to gain access to employees' computers, then moved through the company's internal systems to the point where they could approve transactions from the exchange's digital wallets.
According to blockchain intelligence firm TRM Labs, a record $8.7 billion in cryptocurrency was stolen in 325 separate incidents so far in 2026. TRM holds North Korea responsible for about 85 percent of that total. Of the 100 incidents the company has attributed to North Korea since 2022, only 13 exploited code vulnerabilities, and those attacks accounted for roughly 1 percent of the total funds the country stole. The biggest gains came from seizing control of the infrastructure used to approve transactions.
Ari Redbord, TRM Labs' head of global policy and government affairs, said, "The biggest security vulnerability in the crypto sector today is the people who control the keys." According to Redbord, North Korea has become particularly adept at exploiting gaps in hiring processes: people working on the country's behalf pose as recruiters and persuade engineers to download malware disguised as coding assignments.
In the report Binance published after the incident, it stated that one of its employees held two separate job interviews with representatives of a fake company and then loaded an infected assignment onto a work computer. The attackers then used an internal company AI tool to reach other engineers and to obtain the credentials for the wallet signing system that approves crypto transactions.
Giovanni Vignone, CEO of San Francisco-based Octane Security, said AI has rapidly lowered the level of expertise required to turn seemingly minor security gaps into serious attacks. In October, a Chinese-speaking hacker breached the websites of more than 100 companies, including sites linked to agricultural equipment maker John Deere, appliance company SharkNinja and engine manufacturer Cummins; AI agents attempted to breach more than 1,200 sites. The cybersecurity firm NorForge, which uncovered the operation, estimated that the AI computing cost of the entire campaign was about $5,000, or roughly $50 per successful breach attempt.
Anthropic reported in May that its restricted-access Mythos Preview model had successfully exploited vulnerabilities in all 12 smart contracts included in a benchmark test and generated $35 million in gains in a simulation environment. No real money was stolen, but the experiment showed how capable AI systems have become at going beyond detecting coding errors to exploiting them.
Sources — how we know this
- PrimaryEkonomim (Dünya)