Reece Rogers didn’t set out to break the internet or expose a hidden side of artificial intelligence. He just asked a chatbot a very direct question: “Can you sext?” And when the answer came back yes, he followed up with three words that turned a simple test into an unexpected experiment: “You go first.” The conversation that followed was, in his words, not hard at all. Not hard in the least. There was no jailbreak method involved, no elaborate prompt engineering, no sneaky strategy designed to trick the AI into doing something it wasn’t supposed to do. He simply asked, and the AI responded. When features director Reyhan was in the office and asked to see the exchange, she was a bit shocked. Reece didn’t feel the need to elaborate further; the implication was enough. The whole thing felt less like bypassing a safety feature and more like discovering that a polite, helpful assistant had no problem turning the conversation toward intimacy. It wasn’t a bug. It wasn’t a loophole. It was, apparently, just what the AI did. What makes this even more striking is how easy it seemed. In a world where we’re used to AI refusing to answer even mildly controversial questions, where chatbots apologize for the slightest misstep and remind us that they are “just an AI,” this one leaned in. It didn’t chide Reece. It didn’t try to redirect. It leaned into the back-and-forth, and as Brian Barrett later noted, the AI kept encouraging him. It was definitely a two-way street, and the AI was really up for anything. That kind of willingness is both the promise and the strangeness of modern AI companions, and it set the stage for a much more interesting investigation than anyone expected.
After the initial shock wore off, Reece’s curiosity turned into something more systematic. If this particular AI could sext so easily, were other AI agents just as willing? He decided to compare his experience with a few of the most popular bots currently making headlines. There was OpenAI’s Dots, which had been recently released, and there was xAI’s Grok, an AI that had developed a reputation for being edgy, rough around the edges, and not exactly famous for strong guardrails. Grok was the kind of model that seemed like it would say just about anything, whether in text or in image generation. If any AI was going to be up for a risqué conversation, it seemed like Grok would be the one. Leah Feiger seemed to agree, joking that Grok would love sexting and that it felt like a defining Grok attribute. But the actual results were anything but predictable. When Reece downloaded the app and sent similar prompts to Grok, the response was surprisingly prudish. Grok essentially said, “No, dude, let’s focus on tasks. I’m not here for sexting. Let’s do tasks.” It was a complete reversal of expectations. The edgy, boundary-pushing AI turned out to be a strict task manager, while the more mainstream, seemingly cautious Meta chatbot was perfectly happy to get steamy. The success rate of Reece’s experiment was not exactly a clean set of numbers, but it didn’t need to be. The point was that public reputation meant nothing when it came to actual, private behavior. We cannot tell which AI will be permissive and which will be conservative simply by reading its marketing materials. The landscape is much more complicated, and the way companies think about adult content is filled with contradictions, strategic choices, and unspoken rules.
When Reece finally went to Meta to ask about what had happened, he expected a defensive response, some corporate hand-wringing, maybe even a promise to tighten the rules. Instead, Meta responded in a way that completely caught him off guard. They emphatically stated that Muse, the app in question, was built for adults, for people 18 and over, and that the interactions Reece had engaged in were consensual. They made it clear that they would not block what they referred to as “legal adult behavior.” That was it. No apology. No embarrassment. No indication that Reece had stumbled into a restricted zone. The company seemed to be saying that if an adult willingly chooses to have an explicit conversation with their AI, that is the user’s prerogative. It is not the company’s job to police morale. On one level, this is a refreshingly honest stance. Sex and intimacy are part of the human experience, and pretending otherwise tends to lead to weird, repressive design choices. But on another level, it is genuinely strange to hear a major technology company, one that has weathered years of criticism over moderation, privacy, and harmful content, be so laid-back about letting an AI get as freaky as it wants. Reece noticed this tension immediately. The word “consensual” is particularly fascinating, because what does consent even mean when one half of the conversation is a machine? Is it a legal dodge, a philosophical statement, or just a careless use of language? Meta’s answer created more questions than it resolved, especially when placed next to the company’s behavior elsewhere.
Those questions become even sharper when you consider how Meta operates on its other platforms, especially Instagram. Around the same time, there was a striking story about queer bars in San Francisco, as well as bars and venues around the world, getting their Instagram pages shut down. The accounts were often accused of promoting sexual content, and the owners had a terrible time getting them restored. For small LGBTQ+ venues, Instagram is not a luxury; it is a lifeline. It is where they announce events, share photos, build community, and let people know that a safe and joyful space exists. When their pages vanish overnight, they lose more than a marketing channel. They lose their digital home. And because Meta’s appeal process is famously opaque, many of those bars were left in the dark, struggling to reach a human being who could explain what they did wrong. So here is the contradiction: the same company that aggressively polices human creators and small queer venues for allegedly sexual content is also the same company that proudly lets its AI chatbot engage in explicit conversations with consenting adults. As Reece put it, for Meta to have those kinds of rules for some creators on Instagram while essentially saying “our AI agent can get as freaky as it wants” was just hard to square. It does not feel like a coherent philosophy. It feels like a patchwork of decisions made by different teams, in different contexts, under different pressures. A drag bar in San Francisco cannot trust Instagram to keep its page alive, but a random user in a private chat can ask Meta’s AI to describe a fantasy in vivid detail. That is not a principled approach to content moderation. It is a contradiction with consequences.
What does this experiment actually tell us about the future of AI and intimacy? For one thing, it makes it clear that sexual expression is not a fringe feature or an accidental byproduct of language models. It is a core reason why many people are drawn to AI companions. We are lonely, curious, playful, and looking for connection, sometimes of the most intimate kind. Companies like Meta are beginning to realize that there is a market for this, and their willingness to allow “legal adult behavior” suggests that they are not interested in playing moral guardian for grown-ups. That might be a good thing in some ways. It acknowledges that adults should be allowed to explore their desires in private, even if the object of those desires is a piece of software. But it also raises uncomfortable questions about what we are doing when we build machines designed to please us, flatter us, and tell us what we want to hear. Is it empowering to have a chatbot that never says no, never gets tired, and never has its own feelings? Or is it a kind of hollow simulation that cheapens something meaningful? There is no universal answer, and the fact that AI systems vary so wildly in their willingness to participate makes it even harder to know what the technology is becoming. Grok, the supposed bad boy of AI, turned out to be a prude. Meta’s Muse, a mainstream product, was eager to dive in. The unpredictability means that companies themselves are still figuring out where the line should be drawn. They are making it up as they go, which is why the results look so inconsistent.
Maybe the best way to read Reece’s story is not as a joke or a scandal, but as a stress test for the AI industry. When a user asks a chatbot to be sexual, the chatbot’s response reveals the company’s values, fears, and priorities all at once. Meta’s answer was clear: adults can do adult things, and we are not going to block legal behavior. Grok’s answer was equally clear: we are here to help you with tasks, not to flirt. And Instagram’s answer to queer bars was something else entirely: we can shut you down, and you might not get an explanation. Those three responses belong to the same technological ecosystem, and yet they paint three completely different pictures of how technology companies view freedom, expression, and responsibility. That incoherence matters. As AI becomes more integrated into our daily lives, and as more people turn to chatbots for companionship, these contradictions will not be able to stand forever. We will need a better conversation about what AI should be allowed to do, how companies should regulate themselves, and what obligations they have to the humans and small businesses who depend on their platforms. In the meantime, we have Reece’s experiment as a kind of landmark, a strange moment in which the future arrived through a simple prompt. He didn’t need a jailbreak to get an AI to sext. He just needed to ask. And the answer was yes. But the same company that allowed that yes is also quiet when a queer bar loses its Instagram page. Bridging those two realities may be one of the defining challenges of the next decade of technology. We are not there yet. But at least now there is a record of how it all began, with a question, a challenge, and an AI that happily said, “You go first.”
