“OpenAI's new cybersecurity-focused AI model is more permissive in its usage policies for vulnerability research than standard models like the one powering ChatGPT.”