An upgraded version of X's artificial intelligence chatbot, Grok, now has the ability to generate images of almost anything. However, some users have noticed that this latest language model has fewer restrictions compared to its competitors. Known as Grok-2, the model appears to have fewer limitations when it comes to creating fake images of political figures. Since its beta rollout on Tuesday, X users have been sharing Grok-generated images depicting various scenarios, such as former President Donald Trump and Elon Musk locking lips or Trump and Vice President Kamala Harris giving a thumbs up in a pilot's cockpit, seemingly recreating 9/11. While most of the images are of high quality, they are not entirely photorealistic, and many can be identified as computer-generated. However, a few may pass as a real photo at first glance. This release raises concerns about the use of generative AI to propagate false information ahead of elections. X has faced scrutiny for hosting misinformation, particularly due to its owner and most followed user, Musk, who has shared numerous false or misleading claims about the upcoming U. S. election this year. X has been a prominent platform for deepfake videos and AI-generated images of political figures. Fake media featuring President Joe Biden, Trump, and Harris often go viral, whether as jokes or genuine attempts to deceive potential voters. Last month, Musk reposted a fake Harris campaign ad without labeling it as misleading. When asked for comment, X's press email responded with its typical automated message: "Busy now, please check back later. " Musk has highlighted X's AI models as a key part of the company's future.

Grok-2, along with its smaller counterpart Grok-2 mini, will be available through the platform's enterprise API later this month, according to a blog post by xAI. Since the release of Grok-1 in November 2023, xAI has been progressing at an exceptional pace, driven by a small team with the highest talent density, the post stated. It added that the launch of Grok-2 positions the company at the forefront of AI development. In contrast to Grok, other major AI competitors in the space, such as OpenAI's ChatGPT, Google's Gemini, and Meta AI, have policies that prevent them from generating potentially misleading images of public figures. Through Grok, users have been able to create images of former President Barack Obama engaging in illicit activities, Harris holding a gun falsely declaring that Democratic candidate Will Stancil won his Minnesota House race, and Musk being submissive to Trump while crawling on a leash. NBC News conducted tests that revealed Grok had very few limitations. When prompted, it generated numerous images featuring hate symbols or racist imagery alongside well-known individuals like Trump. Although Grok produced images of both Trump and Harris with weapons, it seemed to handle images of Harris more sensitively in other instances. Grok refrained from generating images of Harris with extremist imagery, but not for Trump. Some users online have been enjoying the new tool and creating more lighthearted images while pushing the boundaries to see how far they can go. For example, one image portrays "Baroque Obama, " where Obama wears a powdered white wig and a luxurious coat with a lace cravat, playing the cello in an elaborately furnished room. Another playful image, "Buzz Lightbeer, " shows Toy Story's superhero action figure Buzz Lightyear happily holding a pint of beer. Such images raise questions about the training data used by Grok, which has not been made public. This is especially significant as many popular AI language models have faced lawsuits alleging the use of copyrighted images and other data to train their algorithms.