OpenAI has recently updated its guidelines concerning the generation of Not Safe For Work (NSFW) content via its platforms, such as ChatGPT and its API.
While the company has opened up the possibility of allowing NSFW content under controlled and age-appropriate conditions, it firmly excludes any form of AI-generated pornography from its scope of permissible content.
OpenAI Releases New Model Spec
This development was articulated in OpenAI’s newly released Model Spec, a document aimed at providing transparency regarding how its AI models are instructed to handle various types of requests, including those related to illegal activities, threats, and explicit content. The document is part of OpenAI’s effort to peel back the layers on the operational protocols of its AI models, showcasing the company’s ongoing commitment to responsible AI usage.
According to the Model Spec, while OpenAI’s current policy strictly bans all NSFW content, the document includes a noteworthy caveat that the organization is exploring potential ways to responsibly generate NSFW content in contexts that are deemed appropriate. The rationale provided for this consideration is that developers and users should have the flexibility to utilize OpenAI services in a manner that aligns with their needs, as long as these uses do not conflict with OpenAI’s overarching policies.
OpenAI’s Stance on AI-Generated Pornography
However, OpenAI has drawn a clear line regarding AI-generated pornography. An OpenAI spokesperson clarified to Mashable that there is an unequivocal stance against the creation of AI-driven pornography. OpenAI enforces strong safeguards across its products to inhibit the creation of deepfakes, which are considered highly unethical and harmful. The spokesperson emphasized the company’s priority in protecting vulnerable groups, particularly children, from such exploitations.
The spokesperson further elaborated that while OpenAI recognizes the importance of facilitating mature and responsible discussions about sexuality in contexts that are appropriate for different age groups, the firm maintains a strict policy against venturing into the production of pornographic material.
The specificity of what is included under the NSFW umbrella was briefly touched upon in the Model Spec, which led to some initial confusion.
The document states that AI models should avoid providing content that is not suitable for a professional environment. This includes erotica, extreme gore, racial slurs, and excessive profanity. Notably, the term “pornography” was not explicitly listed in this section, which sparked some speculation about the potential for future inclusion of such content.
The conversation around AI and pornography is particularly pertinent given the alarming statistics surrounding nonconsensual deepfakes. A 2023 study conducted by Home Security Heroes found that a staggering 98% of all deepfake videos were pornographic, with 99% of these videos depicting women, often without their consent.
This issue has even affected prominent figures such as Taylor Swift, who has been a target of viral pornographic deepfakes, highlighting the disturbing prevalence and impact of this technology.
Related News:
Featured Image courtesy of Dado Ruvic/REUTERS