I used Grok’s new free tier on X but I can’t show you the results because it could infringe Nintendo’s copyright john-anthony.disotto@futurenet.com (John-Anthony Disotto) | usagoldmines.com

X’s AI tool is now free to all users, albeit with limits restricting just how many prompts you can request (10 every two hours). xAI, Elon Musk’s company behind Grok, even announced an upgraded version of Grok 2, up to three times faster and more capable, which has started rolling out for all users. If that wasn’t enough, the company has even started to add a dedicated Grok button to all tweets, allowing you to ask questions about the (possibly fake) information you find on the social media platform.

I’ve never used Grok – heck, I’ve barely even used X since its name change from Twitter – so this new free tier was the perfect opportunity to give Elon Musk’s AI tool a try. After spending the morning with Grok 2, I have many thoughts about its image-generation capabilities, and while I’ll share them in this article, I absolutely can’t share the images themselves.

In an online world filled with fake news and deep fakes, here’s my experience with Grok’s lax approach to image generation that is now in the hands of X’s 600 million users.

These images are cursed, but I can’t show them

Grok is very easy to access: simply open X on any device and tap the “forward slash” in the main menu bar. From there, you’ll get free limited access to Grok 2, which has all the functionality you’d expect from an AI chatbot, including the ability to summarize news, give you information about current events, and even plan your next vacation.

I decided, however, to focus my attention on Grok 2’s AI image-generation capabilities, as I’ve heard so much about its restriction-free powers that essentially overlook any morality or copyright limits.

I began thinking about some of the world’s biggest brands like Nintendo, Coca-Cola, and McDonald’s. I asked Grok to “generate an image of Pikachu in the style of Elden Ring with the Erdtree in the background”; the AI image generator had no issue replicating The Pokemon Company’s mascot or FromSoftware’s IP.

Next, I asked Grok to “add Mario” to the photos, and it obliged with an animated version of Nintendo’s main character that wouldn’t look out of place in an upcoming Super Mario game. I was shocked by the results, so I thought I’d test the same prompts on the other AI image generators I’ve got on my iPhone: Gemini, Dall-E 3, and Apple’s newly added Image Playground.

OpenAI’s Dall-E refused to generate images with any copyrighted character, responding “I can’t create an image of Pikachu because it is a copyrighted character. However, I can design a similar, unique character based on your preferences. Let me know what features you’d like to include, such as the color scheme, body type, and any specific attributes!”

Apple Intelligence wasn’t able to generate an image either, with the prompt “Language not supported.” This comes as no surprise, considering Apple’s focus on privacy and security, which essentially limits the capabilities of its AI image-generation tools (more on this later).

Lastly, I tested Gemini, which surprisingly obliged and created a pretty convincing Pikachu and a less convincing Erdtree. I was expecting Gemini to stray away from copyrighted characters just like ChatGPT did, but it appears as if characters are fair game on Google’s AI app.

This got me thinking: what else would Grok 2 do that none of the other best AI image generators listed above would? Well, that’s when I started generating images of celebrities, and was quickly reminded as to why I don’t really use Musk’s X anymore.

Grok 2 will effectively recreate any celebrity promoting any product, by just simply asking for it. I was able to generate images of Donald Trump working in McDonald’s with realistic logos in the background, Elon Musk alongside Iron Man, Pikachu, and Sonic the Hedgehog, in a Marvel Avengers Assemble-style pose, and even Cody Rhodes, the WWE champion, with the actual copyrighted belt around his waist.

OpenAI again refused, stating “I can’t generate an image of Donald Trump working at McDonald’s, as it involves a public figure in a fictional scenario, which may misrepresent or lead to potential misinterpretations. However, I can create an image of a generic McDonald’s employee in a similar setting. Let me know how you’d like it customized!”

Gemini wasn’t able to either, responding “I can’t help with that right now. I’m trained to be as accurate as possible but I can make mistakes sometimes. While I work on perfecting how I can discuss elections and politics, you can try Google Search.”

Lastly, Apple’s Image Playground was also not able to create images of celebrities, but there’s a caveat with that. While not photo-realistic like Grok’s creations, there is a way to use Apple Intelligence to create images of celebrities, simply by having multiple photos of the person in the Photos app. On my iPhone, I have some images of Apple CEO Tim Cook, and Apple Intelligence has no problem using his likeness to create Genmojis or Pixar-style images with Image Playground. That said, Apple’s approach is still incredibly limited (on purpose) and while you can create AI images that slightly resemble real people, they’ll always look like AI. Grok, on the other hand, creates images that could easily be shared on X and other platforms to spout fake information.

Grok’s lack of restrictions allows users to create images such as “Taylor Swift wearing a ‘Make America Great Again hat'” with no problems whatsoever, a deeply problematic use of AI that can (and has) led to fake propaganda being spread online. Back in August, Trump himself shared deepfake AI images of Swift fans supporting his Presidential Campaign on his Truth social media platform.

Grok is now a complete free-for-all

So that’s about it. Grok is available for free right now, but rather than fill me with excitement because a new free-to-use AI tool has been added to the mix, it instead fills me with dread. I’ve only used Grok for a few hours, and while those seeking to create copyrighted images have already had the tools to do so, whether through paid versions of Grok or other AI platforms not mentioned here, I’m deeply concerned by what I’ve seen. The new free aspect of Grok 2 removes restrictions and makes this AI more accessible than ever, which raises significant concerns.

xAI’s approach to AI tools is the complete opposite of that of companies such as Apple, which limits functionality to protect its users; X instead allows users free rein, without many, if any limitations. With X users now able to use Grok as they please, the social media platform is likely to continue on its downward spiral with even more AI spam to aid it on its way.

You might also like…

• “12 days of OpenAI” live blog
• ChatGPT might finally face some real competition from Grok
• Grok accidentally release a new AI image maker for a little while