OpenAI investigates spicy version of ChatGPT

The popular chatbot ChatGPT from the company OpenAI is notoriously prudish. Suggestive or sexually explicit statements are taboo. The company is considering relaxing those rules.

OpenAI made public a document, Model Spec, that details the behavioral rules imposed on the company’s AI models. OpenAI says it wants to promote public discussion about AI. This is a first draft version.

This sets objectives such as following user instructions and providing useful answers. The chatbots must consider potential benefits and drawbacks for a wide range of stakeholders. Social norms and legislation must be adhered to. The default behavior of the models is discussed. For example, they assume that users have good intentions. The bots should ask clarifying questions if necessary.

Rules are about following lines of command, protecting creators and their rights, respecting privacy, and not generating content that is inappropriate for the workplace. Particularly sexual or suggestive content. However, in a comment, OpenAI said: ‘We are exploring whether we can responsibly enable NSFW content to be generated in age-appropriate contexts via the API (which gives developers access to the models, ed.) and ChatGPT. We look forward to better understanding user and societal expectations regarding model behavior in that area.”

Over the next few weeks, the general public will also be able to provide feedback on the document. OpenAI promises to provide updates on changes to the document.