In the Wake of Anthropic’s Mythos, OpenAI Has a New Cybersecurity Model—and Strategy
…while models explicitly trained and made more permissive for cybersecurity work require more restrictive deployments and appropriate controls. Over the long term, to ensure the ongoing sufficiency of AI safety in cybersecurity…