Anthropic said it has identified and disrupted attempts to use its Claude models for activities that could aid the development of biological weapons, as well as for conventional weapons software, cyber operations, surveillance, and fraud. The California-based AI company’s threat-intelligence report details misuse by suspected state-backed hackers and criminal groups—including actors linked to Russia’s Midnight Blizzard, Iran, and China—and alleges attempts by Chinese AI firms to replicate Claude’s capabilities. The cases span December 2025 to August 2026.
The findings add to a rising drumbeat of warnings about frontier AI risks. Google separately disclosed an effort to use its Gemini model to obtain instructions for synthesizing weaponized biological agents. In Washington, Sen. Bernie Sanders is pushing legislation to ban AI superintelligence and pause advanced AI work, while President Donald Trump has downplayed existential risks in favor of staying competitive. OpenAI Chief Scientist Jakub Pachocki has called for voluntary slowdowns until stronger safeguards are in place. Anthropic said it is tightening controls, sharing intelligence with authorities and peers, and incorporating lessons into its prevention and detection systems.
Related articles:
– AI Risk Management Framework (NIST)
– Guidelines for Secure AI System Development (NCSC)
– OECD AI Principles
