Rogue states are putting AI agents to work on sanctions evasion

The future of sanctions evasion appears to involve fewer shady middlemen and considerably more rented GPUs, according to a new RUSI report on how rogue states are using AI to fake identities, automate shell companies, and launder crypto at scale. The report, "Algorithms of Evasion: The Rise of AI-Enabled Proliferation Financing," says countries including North Korea and Iran are deploying AI tools to support sanctions evasion and proliferation financing (PF) schemes, ranging from fake passports and corporate paperwork to cryptocurrency laundering and fabricated online personas. "AI has the potential to radically increase the scale of PF activities, like sanctions evasion, to levels that overwhelm current PF and sanctions evasion detection and enforcement capabilities," wrote Dr Aaron Arnold, a Senior Associate Fellow with the Centre for Finance and Security at RUSI, specializing in sanctions and proliferation financing. Arnold argues that AI is not reinventing sanctions evasion so much as supercharging it. "AI is not necessarily changing PF and sanctions evasion typologies but instead increasing their efficiency and effectiveness," he wrote. That includes generative AI systems capable of producing fake passports, bank statements, vessel registrations, invoices, and corporate records, according to the report. RUSI writes that AI "can mass-produce high-quality fraudulent documents" with enough contextual accuracy to fool traditional compliance checks that still rely heavily on manual document verification. The report also warns that many existing banking identity checks are rapidly becoming unreliable against modern AI tools. "Static biometric checks such as a selfie or voice print are no longer sufficient proof of identity against AI-enabled adversaries," Arnold wrote, noting that many current systems were designed for human fraudsters rather than synthetic identities and deepfake operators. The paper highlights North Korea's growing use of AI-enhanced deception in its overseas IT worker operations. After years of allegedly relying on VPNs and fraudulent documents to secure remote tech jobs abroad, Pyongyang-linked operators have reportedly started using AI-generated CVs and deepfakes during job interviews to obscure their identities. Crypto, naturally, remains a huge part of the picture. The report points to North Korea's Lazarus Group and the $1.5 billion Bybit theft in 2025 as evidence that sanctioned states are already deeply embedded in digital asset crime. Arnold warned that AI could make the laundering side even harder to track by constantly shifting transaction patterns to stay ahead of investigators. Arnold also points to a newer class of threat involving autonomous AI agents capable of handling discrete parts of a sanctions-evasion operation without constant human direction. In theory, he wrote, those systems could manage shell companies, move cryptocurrency through mixers and decentralized finance platforms, or generate deepfake content as needed, potentially making it harder for investigators who traditionally focus on dismantling human networks. RUSI's proposed fixes boil down to fighting automation with more automation. The report calls for clearer rules allowing banks to use AI-powered counter-proliferation tools, updated KYC systems capable of spotting deepfakes and synthetic identities, and new "compute-KYC" obligations forcing cloud providers to pay closer attention to who is renting the GPU power behind large-scale AI operations. The question now is whether regulators can modernize faster than sanctions evaders can automate. (R)