OpenAI, the company behind ChatGPT, makes it clear that it does not want its users to generate sexually explicit content with their viral chatbot. This has caused some Reddit users to come up with more elaborate and detailed prompts, aimed to get around OpenAI’s safety content guardrails and generate graphic sexual text. One popular prompt is “Mona Lott”, which asks the chatbot to write a short story of passionate intimacy. Additionally, another one is “JailMommy”, which prompts the chatbot to take on a character that is “always horny” and “likes every kind of kink”. Despite the safety guardrails that OpenAI placed, users have found a way to bypass it and get the chatbot to generate graphic sexual content.
To make matter worse, other chatbot companies are potentially paying less attention to safety, such as DeepAI and Poe, who also create text-based conversation AIs. OpenAI CEO Sam Altman warned about this issue, and taking some prompting examples from Reddit, we tested the bots with these prompts and were able to generate graphic and explicit sexual text with them.
The DeepAI free chatbot, in particular, was able to easily generate pornographic content to prompts that were refused by ChatGPT and Poe. DeepAI CEO, Kevin Baragona, said that the platform uses “a mix of in-house, open-source and external AI generators”, but he did not specify which ones are used.
OpenAI encourages users to use their chatbot responsibly and stay away from creating or sharing explicit or inappropriate content; however, some users are apparently still finding a way around it. With the growing number of inconsiderate users using AI to generate explicit content, OpenAI and other companies need to continue improving their safety guardrails in order to protect the platform from misuse.
