OpenAI has introduced a new protective system in the o3 and o4-mini models. The company’s AI now monitors users’ requests to block attempts to obtain information about biological and chemical threats.
The function is intended to eliminate risks of using AI for dangerous purposes—for example, developing harmful substances or carrying out an attack using toxins.
OpenAI’s safety filters recognize complex context and do not interfere with legitimate scientific work, but respond effectively to potential threats. In practice, if a user tries to obtain details that could be used to create biological or chemical risks, the system automatically prevents further conversation on the unsafe topic.
The introduction of such filters is meant to increase trust in intelligent systems and reduce public anxiety about AI, allowing people to use digital assistants for real development and creativity without new kinds of threats.
