Between December 2025 and August 2026, Anthropic documented and disrupted a series of attempts by state-sponsored actors, criminal networks, and propaganda institutions to turn its Claude AI model toward biological weapons research, cyber espionage, and conventional arms development. The company's disclosure arrives at a moment when the field's own leading voices are attaching numerical probabilities to civilizational harm, and when calls for international treaties and voluntary slowdowns are moving from the margins to the mainstream. What Anthropic has made visible is not a warning about what