The gray areas of Midjourney and Stable Diffusion, these mysterious start-ups that create photos with AI

Midjourney CEO David Holz (left) and Stability AI boss Emad Mostaque. Bobby Longoria/Getty Images North America/Jason Henry/The New York Times/Redux/REA

THE pope francis who roams around in an immaculate designer down jacket. An elderly protester, in the hands of the police, his face bloodied. Donald Trump resisting a muscular arrest. These images generated by artificial intelligence have been around the world in recent weeks causing trouble among those who have seen them: true? False? Their increased realism illustrates the great leap forward achieved in just nine months by this technology, capable of generating any type of visual from a simple textual description. Appearing in the summer of 2022 and still in its infancy, these tools were eclipsed in the fall by the ChatGPT phenomenon of the American company OpenAI, champion of automated text generation. The controversies of this spring have brought them back to the fore.

But who is hiding behind these false images? If OpenAI operates the Dall-E visual generation platform, the two most popular services among Internet users and professionals…