Google’s Gemini AI model can once again be used to generate images that represent people, according to Ars Technica. The company had shut down the feature in February after it was criticized for creating historically inaccurate images often related to the subject’s skin color.
The now-updated Gemini 1.5 model should be able to avoid the earlier problems; Google says it also cannot generate images of photo-realistically identifiable individuals, such as famous people.
Google in a blog post Wednesday said it has also focused on improving both AI security and responsibility to ensure the technology is used in an ethical way.