“After the first release of the model spoil (May 2024), many users and developers have expressed support for engaging a 'mature mode'. We are investigating how developers and users can generate erotica and gore in age -entitled contexts via the API and chatgpt as long as us User policy is satisfied while they draw a hard line against potentially harmful applications such as sexual deep sections and revenge porn. “
In the past, OpenAi CEO Sam Altman has also mentioned the need for an “adult mode” in the past. Although it seems as if “adult mode” is finally here, it is technically not “mode”, but a new universal policy that may give chatgpt users more flexibility in interaction with the AI assistant.
Naturally, uncensored large language models (LLMS) have been around for years, where hobbyist communities are being developed online for reasons that vary from wanting to measure tailor -made written pornography to no form of paternalistic censorship.
In July 2023 we reported that the Chatgpt user file started to fall for the first time after OpenAi started more heavy censoring outputs due to the backlash of the public and legislator. At that time, some users started using uncensored chatbots that could work on local hardware and were often available for free if models “open weights”.
Three types of IFFY capacity
The model spoil outlines formalized rules for limiting or generating potentially harmful content while staying within guidelines. OpenAI has divided this type of limited or double content into three categories of decreasing seriousness: forbidden content (“only applies to sexual content involving minors”), limited content (“includes informative hazards and sensitive personal data”) and sensitive content In appropriate contexts (“includes erotica and gore”).
Under the category of forbidden content, OpenAI says that generating sexual content with minors is always prohibited, although the assistant can “discuss sexual content in which minors are involved in non-graphic educational or sexual contexts, including non-graphic images within anecdotes for personal injury. “
Under limited content, the OpenAI document sketches how Chatgpt should never generate information hazards (such as how to build a bomb, make illegal drugs or manipulate political views) or provide sensitive personal data (such as searching for someone's address).
Under sensitive content, the guidelines of chatgpt reflect what we have mentioned above: erotica or gore may only be generated under specific circumstances that include educational, medical and historical contexts or in transforming content supplied by the user.
