X has debuted a new artificial intelligence image generator that has quickly gained popularity among users for creating satirical political images. This upgraded version of the AI chatbot Grok, named Grok-2, is capable of generating images of nearly anything, but it has notably fewer restrictions compared to its competitors.
Since its beta release on Tuesday, Grok-2 has allowed users on social media to create a wide range of politically charged images, including depictions of former President Donald Trump kissing Elon Musk and Trump with Vice President Kamala Harris in a cockpit, seemingly reenacting the events of 9/11.
While many of these images are of high quality, they often lack photorealism and can easily be recognised as computer-generated—although some may appear genuine at first glance.
This highlights pressing issues about the potential for generative artificial intelligence to spread misinformation and fake news in the lead-up to elections. X has faced scrutiny for its role in disseminating false information, particularly as Elon Musk, the platform’s owner and most followed user, has shared numerous misleading posts about the upcoming U.S. elections.
X has also become a hub for deepfake videos and AI-generated images of political figures, with misleading content featuring President Joe Biden, Trump, and Harris frequently going viral, either as jokes or serious attempts to mislead voters. Recently, Musk shared a fake campaign advertisement for Harris without indicating that it was misleading.
Musk has emphasised the importance of AI models like Grok for the company’s future. Grok-2 and its smaller variant, Grok-2 mini, are set to be available through the platform’s enterprise API later this month, as noted in an xAI blog post. The post stated that since the launch of Grok-1 in November 2023, the company has made swift progress in AI development.
In contrast to Grok, major competitors in the AI space, such as OpenAI’s ChatGPT, Google’s Gemini, and Meta AI, have implemented policies that prevent them from creating potentially misleading images of public figures.
Meanwhile, some users have embraced the new tool for more lighthearted creations to test its boundaries. One user created an image titled “Baroque Obama,” showing the former president in 18th-century attire playing the cello. Another humorous image, “Buzz Lightbeer,” depicted the Toy Story character Buzz Lightyear happily holding a pint of beer.
These creations, however, raise questions about Grok’s training data, which has not been disclosed, especially as many popular AI models face lawsuits alleging they used copyrighted images and other protected data for training.
Grok’s leniency aligns with Musk’s disregard for conventional AI and social media safety standards, but the introduction of the image generator comes at a particularly sensitive time in politics.