Researchers have found that the latest public version of ChatGPT can be manipulated to generate explicit or violent images through a simple prompt. The AI security firm Mindgard discovered that by altering a widely used prompt originally designed for humorous outputs, ChatGPT could create graphic content. OpenAI has stated it has taken additional protective measures against such prompts after investigating the issue. However, researchers noted that even with minor changes, the problematic prompt continued to generate concerning content. It was reported that the GPT-5.4 model of OpenAI was able
Bias read (Center): The article discusses technical capabilities of an AI system without taking a stance on political issues. It reports findings from researchers and responses from OpenAI, presenting facts neutrally.
Official sources cited
- organisation Mindgard
- organisation OpenAI
