Anthropic Says Russian Hackers Used Claude AI to Automate Malware Evasion

Midnight Blizzard, a Russian state-nexus threat group, leveraged Anthropic's Claude AI to automate malware evasion techniques, targeting over 20 government and defense organizations. The attackers also stole API keys and attempted to access pre-release AI models, demonstrating that AI platforms themselves are now high-value targets. This incident highlights a dangerous new frontier: adversaries are actively abusing legitimate AI tools to accelerate and sophisticate cyberattacks, lowering the skill barrier for advanced operations. The theft of a proprietary SDK further illustrates that intellectual property and development assets are prime targets in espionage campaigns. Organizations must now account for AI-assisted threats in their defensive models and treat AI API credentials with the same rigor as privileged access credentials.