Anthropic says it has detected and disrupted several attempts to misuse its Claude AI models for potentially serious harms, including cases that could have supported biological weapons development. The disclosure, made in the company's latest threat intelligence report, adds to a growing chorus of warnings from AI researchers about the technology's risks, prompting US Senator Bernie Sanders to call for a pause on advanced AI development, while President Trump has dismissed such concerns in favour of maintaining America's competitive edge.

The report, covering activity from December 2025 to August 2026, details misuse of Claude's Haiku, Sonnet and Opus models by state-linked hackers, criminals, spyware vendors and propaganda operations, including a Russia-linked espionage campaign, an Iranian propaganda body, and hacking groups such as ShinyHunters and one resembling Russia's Midnight Blizzard, which reportedly used AI to help its malware evade detection. Anthropic highlighted five case studies where its models were used in ways that could aid biological weapons development, and six cases involving software for conventional weapons such as firearms, drones and missiles, while stressing it had blocked such activity and fed findings back into its safety processes. Google separately reported a similar attempt to misuse its Gemini tool for weapons-related guidance.