Weaponized AI risk is 'high,' warns OpenAI - here's the plan to stop it

OpenAI is focused on assessing when AI models are sufficiently capable to either help or hinder defenders, and on safeguarding its own models against cybercriminal abuse.