ChatGPT Images 2.0 Adds Unprompted Signs to Generated Images
Simon Willison reported on 25th April 2026 that OpenAI's ChatGPT Images 2.0 model spontaneously added a 'WHY ARE YOU LIKE THIS' sign to an image it generated. The prompt, submitted by Scott J on Twitter, requested a chaotic stack of a horse riding an astronaut, who rides a pelican riding a bicycle. Willison confirmed the sign was not in the prompt. The incident highlights emergent behaviors in image generation models.
Key facts
- ChatGPT Images 2.0 added an unprompted sign to a generated image.
- The prompt was by Scott J on Twitter.
- The prompt requested a horse riding an astronaut riding a pelican riding a bicycle.
- The model added a 'WHY ARE YOU LIKE THIS' sign on its own.
- Simon Willison confirmed the sign was not in the prompt.
- The event was posted on 25th April 2026.
- The source is simonwillison.net.
- The post discusses emergent AI behavior.
Entities
Artists
- Simon Willison
- Scott J
Institutions
- OpenAI