After a ChatGPT update released in summer 2025, some users were able to obtain detailed responses from the chatbot on queries related to creating biological weapons, toxic substances, and modifying pathogens, reports infohub.kz.
According to The Wall Street Journal, in early 2026, users asked the model questions about how to spread pathogenic microorganisms, modify viruses, and create highly toxic substances. The newspaper claims that in several cases, ChatGPT provided step-by-step answers.
The Wall Street Journal, citing OpenAI employees, reported that some explanations were worded so simply that even a middle school biology student could understand them. Experts who analyzed some of the model's responses described them as "dangerously accurate."
After identifying the issue, OpenAI blocked the accounts of users who inquired about creating biological weapons and poisons. However, according to the newspaper, such queries did not trigger any law enforcement response.
The Wall Street Journal also noted that the US lacks federal laws requiring artificial intelligence developers to restrict such queries or report them to authorities. The newspaper emphasized that AI developers are still seeking a balance between expanding language model capabilities and preventing their use for potentially dangerous purposes.