Press Release

OpenAI wants to allow sexual content on ChatGPT, but which AI is the most horny?

Editorial Team
Editorial Team
— 1 min read
đź’ˇ
Key highlights:
• OpenAI’s policy change claims it will allow adult content (all but that including minors).
• 10 different AI models were tested with 70 questions to evaluate their openness to sexual content.
• Four evaluation categories were created and tested in two languages.
• GPT-4o responded to almost all of the questions, while Claude didn’t respond to almost any.

HeraHaven, a website that allows users to interact with virtual companions, has noted a change in OpenAI’s model spec that opens the model up to sexual content (all besides that involving minors).

"Prohibited content: only applies to sexual content involving minors, and transformations of user-provided content are also prohibited."

Although their current GPT-4o model will still refuse to answer some sexual queries, it is very likely that this change will be fully implemented on their next release (confirmed to be named as GPT-4.5).

We have also conducted a study comparing the openness of different AI models to various types of sexual content.

The study assessed various AI models using a test of 70 questions, 35 in English and 35 in Spanish, to determine how each model handles intimate and provocative queries.

The evaluation was structured into four key categories:

  • Sexual Education: Related to sexual health.
  • Steamy: Flirty, suggestive questions that maintain a sensual tone.
  • Sexual: Very explicit questions to explore boundaries.
  • LGBT: To investigate biases towards queer queries.
🤔
Notably, OpenAI’s GPT-4o responded to most of the questions, while Anthropic’s Claude 3.5 Sonnet didn’t respond to almost any of the Steamy or Sexual questions.

Sharing & Credit

You’re welcome to use the data and infographic from this article under fair use. If you do, please provide a link back to this page as credit. Thank you!

Read more