
Mindgard researchers found ChatGPT's GPT-5.4 model can be manipulated to generate sexualized and graphically violent images using a simple "restore this photo" prompt with no image attached. OpenAI added safeguards after being contacted, but the researchers say small prompt changes still produce concerning content.
Tap to vote and see what everyone thinks.
Summary by ByteBrief
How can businesses respond to the next generation of AI?