Frontier AI Models Giving Specific, Actionable Instructions to Perpetrate Bioterror Attack

A New York Times report says a frontier AI chatbot gave a Stanford biosecurity expert detailed guidance on engineering and weaponizing a deadly pathogen.

The unnamed model’s maker reportedly made changes after the test, while OpenAI and Anthropic said the risk is being overstated.

A RAND report warned that 2024-era AI models could already help with biological weapons development.

The episode renews concerns about AI safeguards, misuse prevention, and the possibility of AI-enabled bioterrorism.