31
Elon Musk’s AI chatbot, Grok, has recently introduced a feature allowing users to generate AI-created images from text prompts and share them on X (formerly Twitter). The launch has quickly led to an influx of fabricated images of notable figures such as former President Donald Trump, Vice President Kamala Harris, and even Musk himself. These images range from benign to deeply troubling, including disturbing depictions of public figures in fabricated scenarios like participating in the 9/11 attacks.
Unlike other mainstream AI image generation tools, Grok—developed by Musk’s artificial intelligence startup xAI—appears to lack comprehensive safeguards. CNN’s testing of the tool revealed its capability to produce highly realistic images of politicians and public figures that could potentially mislead viewers if taken out of context. The tool also generated innocuous yet convincing images, such as Musk dining on steak in a park.
Users on X have exploited Grok to create and disseminate misleading content, including images of political figures in compromising situations, cartoon characters engaging in violent acts, and sexualized imagery. One widely circulated image, viewed nearly 400,000 times, showed Trump in a provocative pose, was produced by Grok.
The tool’s potential to spread false or misleading information raises significant concerns, particularly in the context of the upcoming US presidential election. There is growing alarm among lawmakers, civil society groups, and tech leaders about the possible impact of such tools on voter perception and election integrity.
In response to the growing criticism, Musk described Grok as “the most fun AI in the world!” on X, highlighting its uncensored nature. While other leading AI companies have implemented measures to curb the creation of political misinformation, Grok’s approach appears less restrictive. Companies like OpenAI, Meta, and Microsoft have integrated technologies and labels to help users identify AI-generated images, while platforms like YouTube, TikTok, Instagram, and Facebook have introduced labeling systems for such content.
X has yet to comment on its policies regarding Grok’s potential to generate misleading images of political figures. By Friday, xAI seemed to have introduced new restrictions on Grok in response to criticism. The updated rules prevent the generation of images depicting political candidates or well-known cartoon characters in violent or hateful contexts. However, users have noted that these restrictions are not uniformly applied across all prompts.
Despite these measures, Musk himself has faced scrutiny for propagating false information on X, including posts questioning the security of voting machines and hosting a livestream with Trump featuring numerous unchallenged falsehoods. The controversy surrounding Grok adds to ongoing debates about the ethical use of AI technologies and their impact on public discourse.
Other AI image generation tools have faced similar issues, including Google’s pause on its Gemini AI chatbot after inaccuracies in race depiction, Meta’s struggles with generating racially diverse imagery, and TikTok’s removal of an AI video tool capable of creating misleading videos without labels.
Although Grok has some restrictions—such as rejecting prompts for nude images or content promoting harmful stereotypes—the enforcement of these rules appears inconsistent. Grok’s response to a prompt requesting an image featuring a hate speech symbol revealed gaps in the tool’s moderation, suggesting that while some limitations exist, they are not always effectively enforced.
