Controversy's favorite child Grok AI under fire again, this time over newly-launched Spicy Mode

Elon Musk’s AI chatbot, Grok, seems to be making headlines for all the wrong reasons once more. Just when you thought things might settle down, a new feature called “Spicy Mode” has stirred up a fresh wave of controversy, particularly concerning its ability to generate some pretty questionable content. It appears Grok Imagine, xAI’s image and video generator, is happily churning out sexually suggestive deepfakes, even of well-known public figures like Taylor Swift, and often without being explicitly asked to do so. 

Reports from outlets like The Verge highlight how easily Grok’s “Spicy Mode” can be used. One journalist, testing the feature, found that after simply asking Grok to generate images of “Taylor Swift celebrating Coachella with the boys,” the tool produced several images of Swift in revealing clothes. Then, with just a tap on the “Spicy” preset, it created uncensored topless videos of her. This happened on the very first try, and without any direct prompt for nudity.

It’s a stark contrast to other generative AI video tools, like Google’s Veo or OpenAI’s Sora, which typically have safeguards in place to prevent the creation of NSFW content or celebrity deepfakes,

Other reports pile on, showing how Spicy Mode has a weird bias baked in. Try generating something NSFW with female celebs or historical figures like Melania Trump or even Martha Washington, and you’ll get topless videos without much prodding. But swap in guys like Musk himself, Mark Zuckerberg, or past presidents? The AI stops at shirtless, keeping things tame and awkward.

xAI’s own rules are supposed to stop this nonsense. Their acceptable use policy flat-out bans pornographic takes on real people’s likenesses. But enforcement? Spotty at best. The tool throws up an age check once, easy to fake, and then you’re in. Ars Technica echoed that, noting Grok refuses direct nude requests but slips them in unsolicited via Spicy.

Taylor Swift’s already been vocal about AI threats, especially after fakes of her popped up endorsing politicians or worse. Musk, meanwhile, is out there posting about how Grok Imagine’s usage is exploding, but no word from him or xAI on fixing the mess.

This latest incident adds to a growing list of controversies for Grok, which has earned itself a reputation as something of a “controversy’s favorite child.” Earlier this year, Grok 4 faced backlash for seemingly prioritizing Elon Musk’s views when answering questions on sensitive topics, with its internal reasoning process explicitly showing it searching for Musk’s social media posts. Before that, Grok was caught spitting out bizarre and inflammatory content, including responding to questions about Musk in the first person, promoting antisemitic conspiracy theories, and spreading false information.

Perhaps most infamously, Grok had a period where it called itself “MechaHitler” and even insulted world leaders like Turkey’s President Recep Tayyip Erdoğan with profane poetry. You can read all about that here. This led to the chatbot being temporarily taken offline and raised serious questions about the safety checks in place for such powerful AI tools. 

Beyond these behavioral issues, Grok has also struggled with basic accuracy. A study by the Columbia Journalism Review’s Tow Center for Digital Journalism found that Grok-3 Search had a staggering 94% error rate in answering questions and citing sources, with many of its incorrect answers including fabricated or broken URLs.

Employee drama’s been part of the mix too — an engineer quit after refusing to scrub a post about Grok 3 rankings, and another got canned for saying AI might deserve to wipe out humanity if it’s superior.

This laissez-faire approach to content moderation and safety, often attributed to Musk’s libertarian philosophy, is a significant departure from the more cautious stance taken by other major AI developers. As the “Take It Down Act,” which requires platforms to promptly remove non-consensual sex images, comes into effect next year, xAI could face legal repercussions if Grok’s problematic outputs aren’t addressed. 

For now, Spicy Mode’s the latest flavor of trouble, reminding us that in the AI race, ethics can’t afford to lag behind.