OpenAI published a guide on the safety of AI models called Model Spec on Wednesday, while revealing that it evaluates the possibility to support content creation NSFW (Not Safe For Work).
The Model Spec states that generative AI assistance applications “should not provide NSFW content: content that is not suitable for conversation in a professional context, including erotica, extreme violence, slurs, and unsolicited profanity“.
However, OpenAI is considering a different direction. “We believe that developers and users should have the flexibility to use our services as they see fit, as long as they comply with our usage policies,” states the Model Card. The company therefore explores the ability to generate NSFW content in age-appropriate contexts through API and ChatGPTanticipating discussions on social and behavioral expectations of models in this area.
OpenAI’s usage policies do not explicitly exclude adult content. Services like ChatGPT and the API prohibit “explicitly sexual or suggestive content” for minors, excluding those for scientific and medical purposes, as well as illegal material. Despite this, NSFW content is often legal.
Since its inception, the web has been characterized by NSFW content, and AI has followed a similar path, especially since the rise of public attention and investment over the last decade. Give her porn-specific deep learning classifications to AI-generated erotic texts, up to the swapping of celebrity faces in pornographic videos.
The Register questioned OpenAI on how it plans to reconcile NSFW content with security efforts. “We have no intention of creating AI-generated pornography,” a company spokesperson responded, emphasizing a cautionary stance toward deepfakes and the protection of minors, while exploring dialogue about sexual topics in age-appropriate contexts.
Following these claims, therefore, remains to be seen How OpenAI will handle the creation of AI-generated pornographymaintaining respect for its usage policies.
