Yesterday, I reported that Meta's AI image generator was consistently depicting everyone as Asian, even when the text prompts clearly indicated a different race. However, today I encountered the opposite issue: I could not generate any images of Asian people using the same prompts as before. My tests yesterday involved using the AI image generator available in direct messages on Instagram. Despite numerous attempts, I could not generate a single accurate image when using prompts such as "Asian man and Caucasian friend" or "Asian man and white wife. " The system would consistently depict everyone as Asian, with only one exception where it successfully depicted an Asian woman and a white man. After reaching out for comment yesterday, a Meta spokesperson requested additional details about my story, including my deadline. I promptly responded but did not receive any further communication. Today, I wanted to determine if the issue had been resolved or if the system still couldn't accurately generate images of Asian individuals with their white friends. However, instead of incorrect images, I received an error message stating, "Looks like something went wrong. Please try again later or try a different prompt. " Curiously, I wondered if I had reached a limit for generating fake Asian people. To investigate further, I asked a colleague from The Verge to attempt the same process, and she encountered the same result. I also attempted other, more general prompts related to Asian people, such as "Asian man in suit, " "Asian woman shopping, " and "Asian woman smiling. " Unfortunately, I consistently received the same error message instead of an actual image. Once again, I reached out to Meta's communications team to inquire about this issue. However, as time passed, I still hadn't received a response from Meta, even after forty minutes and the conclusion of a meeting.

Interestingly, during this time, the Instagram feature began working for simpler prompts like "Asian man. " It is not uncommon for companies I cover to silently modify or remove a feature after a reporter brings attention to it. Could this be the case here?Did my inquiry cause a temporary shortage of AI-generated Asian people?Or was it merely a coincidence in timing?I wish I had answers, but Meta never provided any explanation or responded to my questions. Whatever may be happening within Meta's headquarters, it's clear that there is still work to be done. While prompts like "Asian man and white woman" now generate an image, the system still fails to accurately depict their races and instead makes both individuals appear Asian, similar to what occurred yesterday. It seems we are back to where we started. I will continue to monitor the situation closely.