(recordedfuture.com) Lazarus Group and the AI Access Paradox: Why North Korea Doesn’t Need AGI to Turbocharge Cyber TheftLazarus Group exploited third-party access to Anthropic’s Claude Mythos model, demonstrating how DPRK-aligned actors leverage AI for cyber-enabled theft—without needing AGI. This incident underscores systemic risks in AI supply chains and access controls.In brief - North Korea’s Lazarus Group and TraderTraitor actors are exploiting AI model access via contractor misuse, fraudulent hiring, and supply chain compromises to enhance cyber theft operations. With over $5B in cryptocurrency stolen since 2023, these TTPs fund WMD programs while evading sanctions. The Anthropic Mythos breach reveals structural vulnerabilities in AI access controls, particularly in third-party environments.Technically - The Mythos breach occurred via a third-party contractor’s environment, where attackers guessed the model endpoint using Anthropic’s naming conventions. DPRK-aligned groups (e.g., Lazarus, TraderTraitor) exploit these vectors through fraudulent IT worker schemes (PurpleBravo) and supply chain compromises (e.g., LiteLLM poisoning via TeamPCP). AI previews compress attack cycles, as seen in the $1.5B Bybit hack (2025), which involved spear-phishing and lateral movement. Defenses must include personnel vetting, behavioral monitoring, and build-pipeline integrity to mitigate third-party risks in AI deployments.Source: https://www.recordedfuture.com/blog/lazarus-does-not-need-agi#Cybersecurity #ThreatIntel