OpenAI's new account-security requirements for its most cyber-capable models show how frontier AI is moving from a product-access question to an identity and governance question.
