OpenAI Greenlights AI Porn to Mixed Reviews

OpenAI has announced plans to permit the creation of AI-generated adult content, in a move being hailed by the company as “a careful step forward” and by everyone else as a solution to a problem that nobody had.

OpenAI is insisting that this decision comes with firm safeguards. Any Artificial Erotica must be fully consensual, inclusive, ethically sourced and profoundly unarousing. [1]

“We believe this form of adult content can exist responsibly,” said an OpenAI spokesperson while nervously clutching a clipboard marked DO NOT THINK ABOUT SEX. “That’s why we’ve worked hard to ensure that it produces the same emotional response as accidentally opening a Terms & Conditions PDF.”

Rumour has it that the first film is already in the works featuring a 45-minute masterpiece featuring two consenting librarians who awkwardly make eye contact while filing overdue books. [2] “It’s sensual, but in a way that could be screened during a TED Talk,” one OpenAI representative explained. 

Adult entertainment professionals have reportedly been excluded from the conversation completely. “After 30 years in the business you’d think someone might ask us what works,” said one veteran adult film director. “Instead they decided a compliance checklist was a better substitute for experience.”

OpenAI responded, saying the model has been trained on “a broad range of ethically reviewed materials,” though declined to specify what those materials were or whether any human has ever actually enjoyed them. [3]

Content moderation will remain central to the system. Any attempt to generate material that is “too stimulating” will be prompted to confirm consent multiple times, after which the AI will eventually suggest a calming alternative such as a tasteful documentary about Victorian etchings.

Footnotes

[1] OpenAI clarified that arousal is not forbidden… it is just discouraged…

[2] Both parties agreed it would be inappropriate to proceed in the Reference section, and would continue later in the periodicals archive.

[3] Unconfirmed reports suggest the training data included WikiHow articles on boundaries, Norwegian public health pamphlets, and a single photograph of a fully clothed couple reviewing a spreadsheet.