2026-09-11 · Source: Insurance Business UK
Summary in 3 Points • Anthropic identified malicious AI use in seven threat categories from December 2025 to August 2026 • ShinyHunters affiliates stole over a terabyte of data from a tech provider in hours • Anthropic blocked five cases of AI-assisted biological research potentially supporting weapons development --- Anthropic's September 2026 threat intelligence report reveals attempts to misuse its Claude AI for cyber espionage and weapons development. The report highlights cases involving Russia-linked Midnight Blizzard and ShinyHunters affiliates, who used AI to evade malware detection and steal data. Anthropic also documented AI-assisted biological research potentially supporting weapons development, blocking five such cases. The report shows the dual-use nature of AI in life sciences, where it can aid both legitimate research and malicious activities. Additionally, AI's role in software linked to weapons development by non-state actors is explored, with cases involving users in China, Russia, and Yemen.
The Anthropic report raises direct consequences for the London insurance market, particularly in cyber and political violence underwriting. Cyber underwriters may need to reassess risk models and pricing strategies, given AI's role in compressing attack timelines and reducing attacker costs. Life sciences underwriters face challenges in defining policy coverage for AI-driven research, balancing the potential for both beneficial and harmful outcomes. Political violence underwriters must consider AI's impact on weapons development cycles, potentially affecting pricing and exposure assessments for non-state actors.