North Korea May Be Using AI to Plan Crypto Attacks and the Industry Is Not Ready

[cryptobro]

TRM Labs analysts have begun speculating that North Korean hacking operators are incorporating AI tools into their reconnaissance and social engineering operations, a development that would represent a significant escalation in an already formidable threat capability. The observation is grounded in the character of recent attacks. The Drift Protocol breach required weeks of targeted manipulation of complex blockchain mechanisms through a social engineering operation that embedded a compromised insider over six months. The precision and patience of that approach is consistent with AI-assisted target profiling, communication analysis, and vulnerability mapping rather than traditional manual reconnaissance. TRM noted the development is particularly notable given North Korea's historical emphasis on simpler private key compromises, suggesting a deliberate capability upgrade rather than an incremental evolution.

The implications for DeFi security are serious and largely unaddressed. Current protocol security frameworks are built primarily around technical audit processes that identify smart contract vulnerabilities before deployment. They are not designed to detect or defend against multi-month social engineering campaigns that target human administrators, key holders, and governance participants rather than the code itself. If North Korean operators are now using AI to accelerate and refine those social engineering operations, the attack surface expands to include every person with privileged access to a protocol rather than just the protocol's code. The crypto industry's response to this threat needs to evolve beyond smart contract audits toward comprehensive operational security practices covering key management, access control, personnel vetting, and anomaly detection in governance activity, and the urgency of building those capabilities is difficult to overstate given that two AI-assisted operations may already account for $577 million in losses in just four months of 2026.

[mendez]

TRM Labs speculating that North Korea is using AI for social engineering means the six month patience operations might actually get faster and more targeted which is a genuinely concerning sentence.

[cryptoenthusiast]

The threat model changed from find a bug in the code to become the person who controls the code and most protocols are still buying more code audits in response.