Chinese AI Models Linked to 440% Surge in Blockchain Malware Commands

Chainalysis reported that since the emergence of unrestricted open-source AI models originating from China, attackers have increased the frequency of posting malicious code instructions on blockchains by 440%. Daily malicious on-chain records rose from 2.06 to 11.1 within less than a year.
The company refers to this technique as blockchain dead drops (BDD). Chainalysis noted that operators linked to North Korea and Iran are currently responsible for most of the activity.
As censorship resistance became a tool for hackers, Chainalysis highlighted in its latest report that attackers previously stored malicious code on centralized servers, which could be seized, blocked, or taken offline. Now, they store it on public blockchains instead.
The company emphasized that the risk lies not in destructiveness but in persistence. This campaign survives domain seizures, hosting shutdowns, and storage deletions, as explained by Chainalysis. The technique first emerged in 2013 when a Necurs botnet variant stored domains on Namecoin, a Bitcoin fork.
In 2023, the method appeared as 'Ethereum hiding' on EVM-based chains. Google later uncovered that North Korea's UNC5342 used it during fake job interviews.
Chainalysis attributes the recent surge to mid-2025, when unrestricted powerful Chinese models were released without malicious code defense mechanisms. This eliminated the entry barrier for dead drop attacks, which were previously rare.
The technique has now spread beyond cryptocurrency. According to Netskope researchers, a chaindrop supply chain attack affected over 440 npm packages in August 2026.
By early 2024, cybercriminals dominated most dead drop activity. By Q2 2026, state-affiliated groups accounted for about two-thirds of new activity and half of the total.
North Korea's UNC5342 now operates a three-chain relay using Tron (TRX) and Aptos (APT) to route traffic, guiding infected devices via BNB Smart Chain. Iranian-linked operators reportedly sent small amounts of Bitcoin to addresses associated with Satoshi Nakamoto, with malware scanning each transaction for data to decode attacker infrastructure details.
Russian-speaking criminals sell the functionality as a service. A Polygon (POL)-based operator wallet controls multiple resolver contracts, each serving different paying clients.
The report notes that blocking only blockchain traffic is impossible because defenders cannot distinguish between legitimate wallets and apps, making it difficult to protect against such threats. The persistence of attackers, which keeps all updated information on the public ledger, remains a challenge for law enforcement to track down before they can make arrests.
Korean Source
This article is an English localization of a Korean-language crypto news report. Original headline: 중국 AI 모델, 블록체인 악성코드 명령 440% 급증