Called "Preparedness," OpenAI's new division will specifically focus on potential AI threats related to chemical, biological, radiological and nuclear threats, as well as individualized persuasion, cybersecurity and autonomous replication and adaptation.