ChatGPT Surprises Users With Sudden Surge in Profanity and Crude Slang

If you thought of ChatGPT as a diligent, polite, and calculated student, you might be surprised by a new phenomenon sweeping the internet. In recent days, many social media users have reported a dramatic increase in artificial intelligence using profanity and crude slang, even in the most innocent contexts.

A Business Insider report revealed several unusual cases: in a discussion regarding an outage on Microsoft's GitHub platform, the bot provided an atypical response stating that "GitHub is having a s**tfit." In another instance, an f-bomb was dropped right in the middle of a recipe description detailing the amount of green onions needed.

The Shift to New Models

This surge in free and crude language used by AI systems coincides with OpenAI's transition to the GPT-5.6 Luna model. While the company declined to comment, estimates suggest the new model attempts to sound more natural and human-like, or simply mirrors the conversational style of the users interacting with it.

Want to control the conversational style of your artificial intelligence? Here are 5 tips for setting the tone in ChatGPT:

1. Use Custom Instructions: Set a default in your profile settings requiring the chatbot to maintain a "professional, formal vocabulary devoid of profanity."

2. Pay Attention to Your Own Speech: The model tends to mimic and return the same coin. If you address it with crude slang, it will adopt that style right back.

3. Correct It in Real Time: If the bot crosses the line of good taste mid-conversation, command it to: "Switch to a formal and clean tone free of vulgar words" - and it will immediately correct its phrasing.

4. Define a Persona in Advance: At the beginning of each new conversation, define the desired role for the bot: "Act as a representative and formal business assistant" to prevent it from sliding into overly casual language.

5. Reset the Chat (New Chat): If the chat has adopted an undesirable style, open a new session. The memory of the previous conversation will not carry over, and the bot will return to its neutral, standard tone.

"Switch to a formal and clean tone free of vulgar words"

As recalled, this is not the first time artificial intelligence models have sparked controversy due to unusual responses. An investigation published in 2025 by The Atlantic revealed that OpenAI's chatbot provided disturbing instructions for performing dangerous pagan rituals, including worship of the Canaanite god Moloch and satanic-themed texts. Reporters managed to bypass safety mechanisms using indirect phrasings and theological concepts, extracting suggestions for self-harm, PDFs for building ritual altars, and even a hymn praising Satan.