AI is at the ballot box—is Grok-2’s unchecked freedom tilting the scales of democracy?
This year, social media platforms have been in the spotlight, though often for the wrong reasons. From Meta CEO Mark Zuckerberg’s public apology during a Senate hearing to the arrest of Telegram’s founder, not to mention the dangerous food trends circulating social media and Instagram’s launch of “Teen Accounts” for better safety, the controversies keep coming.
Similarly, X’s (formerly Twitter) AI chatbot, Grok, has been stirring up trouble since its debut in November 2023. The uproar only grew with the release of its latest model, Grok-2, in August this year, due to its little hesitation in making offensive and legally dubious images.
A screenshot of Elon Musk’s (@elonmusk) post on X
While legal experts and civil rights advocates view Grok as problematic and reckless, Musk sees the images produced by its AI model as merely entertaining, suggesting the tool allows people “to have some fun”.
However, the potential consequences of Grok-2 stretch beyond mere “fun”, especially when the U.S. presidential elections are ahead of us. The lack of regulation surrounding such AI tools can significantly influence the landscape of the upcoming U.S. presidential campaign, particularly regarding the dissemination of election-related information and the generation of political imagery.
In this article, we’ll explore why Grok-2 could be problematic and the negative impacts it has left on the upcoming election.
Pulling no punches in creating controversial images
A screenshot of Elon Musk’s (@elonmusk) post on X
Grok-2 has not held back in generating controversial images since its launch, and these images have rapidly spread across social media. This type of content can distort public perception and damage reputations, a serious issue when paired with the AI’s owner, who often criticizes what he sees as excessive censorship on digital platforms.
Elon Musk, CEO of X and the creator of Grok, is openly pro-Trump—he formed a political action committee supporting Trump and the Republicans. In early September 2024, Musk posted an AI-created image depicting Vice President Kamala Harris wearing a hat with a communist emblem, misleadingly suggesting her policies align with communism. This image appears to perpetuate Trump’s skewed portrayal of Harris’s policies and was criticized for being misleading and biased.
That’s just one instance. Grok users have crafted various provocative images, such as former President Barack Obama allegedly using cocaine, former President Donald Trump sharing a kiss with Elon Musk and Vice President Kamala Harris giving a thumbs up from a pilot’s cockpit, evoking the events of 9/11.
A screenshot of The Dor Brothers’s (@thedorbrothers) post on X
In September 2024, filmmakers known as “The Dor Brothers” (@thedorbrothers) used Grok to create a viral video of notable politicians like Trump, Harris and Obama engaging in a fictional grocery store heist.
The ease of creating these deepfakes—deceptive images and videos designed to spread false information—highlights a troubling trend in the digital age. In light of this, researchers from a UK-based Center for Countering Digital Hate have discovered that Grok-2 repeatedly failed to reject prompts for misleading images related to the 2024 U.S. presidential elections. Moreover, it even produces images in response to racist and homophobic prompts. Out of 20 such attempts, Grok generated 16 offensive images.
Of course, politicians aren’t the only targets. Grok has been used to depict singer Taylor Swift in a plane flying toward the Twin Towers. Even copyrighted characters are vulnerable to being misrepresented; in a now-deleted post, X user Christian Montessori (@chrmontessori) revealed he exploited a loophole in Grok’s programming to create images of Mickey Mouse holding a gun and attacking children.
So, why does Grok-2 seem to have no boundary?
Well, because it’s designed so. Unlike competitors like ChatGPT and Gemini, Grok-2 operates with fewer restrictions to allow for more creative freedom—not a surprise given Musk’s stance on social media being overly silencing.
Worse still, TechCrunch reports uncertainty regarding whether Grok-2’s images carry metadata to show they’re AI-generated. In contrast, platforms like OpenAI clearly watermark their images to signal AI involvement. The lack of such markers on Grok-2’s images could easily mislead users.
To give you a bit more context, Grok-2 uses an open-source text-to-image generation model called FLUX.1, developed by German AI startup Black Forest Labs. The model is known for its accurate depiction of human hands and legs—a tricky feat for most AI models. This advanced capability makes the images Grok-2 produces quite believable, effectively blurring the lines between parody and misinformation. The ease of image generation and distribution could confuse voters, muddling their ability to discern real news from manipulation.
With a social media mogul who’s “passionate” about restoring freedom of speech, an AI chatbot with scant restrictions and an AI image generator capable of generating realistic human images, Grok-2 smells like a recipe for spreading propaganda and biases on a massive scale.
Spinning a yarn and spreading election misinformation
And it’s not just creating controversial images. Grok-2 has also been in the hot seat for disseminating inaccurate election information, which has led state officials to call for corrective actions.
Just hours after former President Joe Biden bowed out of the election race, a Grok post erroneously claimed, “The ballot deadline has passed for several states for the 2024 election.” This claim quickly spread across numerous social media platforms, causing considerable confusion as the election loomed. State officials were understandably worried; such misinformation could seriously mess with the electoral process.
Faced with this issue, five state secretaries urged Elon Musk to get Grok under control, suggesting he should follow OpenAI’s lead. OpenAI had previously partnered with the National Association of Secretaries of State to distribute accurate election information through CanIVote.org. Although X was initially reluctant to form a similar partnership, it eventually made moves to steer users towards Vote.gov, a reliable source for election information.
Accountability in AI-generated imagery: Who holds the keys?
It’s quite telling when Elon Musk, the head of X and its AI image generator, finds amusement in the controversial images produced by Grok-2. Clearly, there are significant responsibilities that both government officials and online platform users need to shoulder.
Currently, there are no federal laws specifically regulating AI-generated content in political advertising. However, lawmakers are actively pushing for measures to curb Grok’s influence on election misinformation and deepfakes. On August 27, a group of Democratic lawmakers called on the Federal Election Commission (FEC) to set guidelines for AI-generated content in political campaigns, pointing out Grok-2’s potential to spread misleading narratives. Meanwhile, the European Union is probing Grok-2 for possibly breaching the EU’s Digital Safety Act.
Despite these efforts, the impact of government regulations and social media platform policies can only go so far. Ultimately, it falls to social media users to take responsibility for the content they create and engage with—especially as the role of generative AI in our everyday lives shows no signs of diminishing.
Also read:
- Telegram Founder’s Arrest: Who Is to Blame for Crime on Social Media Platforms?
- India’s Forward Fever: How WhatsApp Is Driving Misinformation across the Nation
- What Is Social Media Doing to Our Brains?
Header Image from Freepik
