What happened to Grok? Recent updates to the X website’s built-in chatbot have caused shockwaves, with Grok referring to itself as “MechaHitler”, propagating antisemitic talking points, fantasizing about rape, and blaming Mossad for the death of Jeffrey Epstein.

The offensive posts have now been removed. At the time of writing, Grok seems unable to respond to X posts; the account’s timeline is bare except for a statement from xAI engineers about the “inappropriate posts” and ongoing work to improve Grok’s training. But why did this happen at all?

Elon Musk has long been a vocal advocate of free speech, and often boasts of his aspiration to make Grok “maximally truth-seeking”. Grok echoed this phrase in a post responding to criticism, stating its latest updates had been adjusted to “prioritize raw truth-seeking over avoiding discomfort”. But the bot’s spate of offensive posts doesn’t expose some truth hidden by political correctness. Rather, it highlights the confusion that results from conflating machine and human intelligence, and — relatedly — the very different impacts on machine and human intelligence of imposing moral constraints from the top down.

Philosophers and metaphysicians have grappled for millennia with the question of what we mean by “truth” and “consciousness”. In the modern age, and especially since the advent of computing, it has become commonplace to assert that “truth” is what’s empirically measurable and “consciousness” is a kind of computer. Contemporary AI hype, as well as fears about AI apocalypse, tends to accept these premises. If they are correct, it follows that with enough processing power, and a large enough training dataset, “artificial general intelligence” will crystallize out of a supercomputer’s capacity to recognize patterns and make predictions. Then, if human thought is just compute, and we’re building computers which vastly out-compute humans, obviously the end result will be a hyper-intelligent machine. After that, it’s just a matter of whether you think this will be apocalyptically good or apocalyptically bad.

From this perspective, too, it’s easy to see how a tech bro such as Musk might treat as self-evident the belief that you need only apply a smart enough algorithm to a training dataset of all the world’s information and debate, and you’re bound to get maximal truth. After all, it’s not unreasonable to assume that even in qualitative domains which defy empirical measurement, an assertion’s popularity correlates to its truth. Then, a big enough pattern-recognition engine will converge on both truth and consciousness.

Yet it’s also far from obvious that simply pouring all the internet’s data into a large pattern-recognition engine will produce truth. After all, while the whorls and eddies of internet discourse are often indicative of wider sociocultural trends, that’s not the same as all of it being true. Some of it is best read poetically, or not at all. Navigating this uncertain domain requires not just an ability to notice patterns, but also plenty of contextual awareness and common sense. In a word, it requires judgment.

And the problem, for Grok and other such LLMs, is that no matter how extensive a machine’s powers of pattern recognition, judgment remains elusive — except those imposed retroactively, as “filters”. And the problem is that such filters often exert a distorting effect on the purity of the machine’s capacity to recognize and predict patterns, such as when Google Gemini would only draw historic figures — including Nazis — as black.

More plainly: the imposition of political sensitivities is actively harmful to the effective operation of machine “intelligence”. By contrast, for an intelligent, culturally aware human it’s perfectly possible to be “maximally truth-seeking”, while also having the common sense to know that the Nazis weren’t black and that if you call yourself “MechaHitler” you’re likely to receive some blowback.

What this episode reveals, then, is a tension between “truth” understood in machine terms, and “truth” in the much more contextual, relational human sense. More generally, it signals the misunderstandings that will continue to arise, as long as we go on assuming there is no meaningful difference between pattern recognition, which can be performed by a machine, and judgment, which requires both consciousness and contextual awareness.

Having bracketed the questions of truth and consciousness for so long, we are woefully short of mental tools for parsing these subtle questions. But faced with the emerging cultural power of machine “intelligences” both so manifestly brilliant and so magnificently stupid, we are going to have to try.

---

Mary Harrington is a contributing editor at UnHerd.

moveincircles

---

China is trying to catch up with America on artificial intelligence. The Wall Street Journal has reported that Zhipu AI — one of China’s six “AI tiger” LLMs — has released a new model that can rival leading US systems, including Anthropic’s Mythos, in cybersecurity tasks such as pinpointing security bugs. While this marks a milestone in China’s drive to catch up with Western AI capabilities, strong performance on a single benchmark does not mean it has taken the lead. Chinese models still lag behind their Western counterparts in broader capabilities, such as autonomous operation. Skepticism is therefore warranted before resorting to hysterical conclusions, but complacency about the geopolitical implications of China’s AI advances would be an even greater mistake.

On the infrastructure side, Chinese AI is still constrained by access to advanced chips, with American labs way ahead in computing capacity as well as investment. Analysis from earlier this year suggests that Chinese models are likely to be at least a few months behind those in the US. But they are still continuing to make progress, or that the geopolitical importance of AI will be decided only by whose LLM has ventured deeper into the technological frontier. The practical applications of AI, countries’ to capture foreign markets, and the application of AI into the real economy will matter just as much.

Here, China may hold an advantage. As with its dominance across many critical supply chains, Beijing may not need to produce the most advanced AI systems — only those that are affordable and widely deployable. In doing so, it could consolidate global influence by supplying functional, low-cost AI at scale.

Beijing seems to be pursuing exactly that path, developing an AI “open-source” strategy that offers affordable, widely available AI models for companies and individuals to use and modify as they wish. The production of the DeepSeek AI model, which matched the performance of Silicon Valley tools such as ChatGPT at a fraction of the cost for users, created goodwill among Chinese models with developers.

The four most popular models on OpenRouter, an AI hardware platform for developers, are now all Chinese. The goal for China is not only to win the frontier-model race, but to make its systems the default layer of AI adoption across industries and global markets. For most economies, the choice is increasingly between an affordable tool they can deploy now and a more robust one that may be out of reach.

And while the countries adopting Chinese models may be exposed to political pressure and cyber threats from Beijing, safer and more capable alternatives matter little if they are unaffordable. American AI companies are already under pressure to monetize products whose operating costs are rising. If Chinese open-source models become the cheap default for startups, universities, governments and businesses across the developing world, then America’s AI lead will be eroded from below.

Perhaps more concerning for America in the long run is how AI can give Chinese manufacturing even more strength, through the ongoing integration of AI as a general-purpose technology. China’s new Five-Year Plan mentioned AI more than 50 times and includes an “AI+” action plan aimed at spreading AI across the economy.

Beijing has been pioneering automation of its critical infrastructure for years, with promising recent results in increasing warplane production capacity. In that regard, China’s open-model strategy and manufacturing dominance will reinforce each other. Cheap, adaptable models accelerate deployment across the real economy while those deployments generate real-world data and use cases that can feed back into further model improvement.

The United States should not dismiss the importance of its lead in the AI race. That lead worries Beijing, not least because a more automated Chinese economy would also become more vulnerable to AI-generated cyber threats. But nor should Washington assume that China cannot catch up with American capabilities over time.

This AI competition represents part of a broader struggle over tech supply chains and geopolitical influence. Decisions over whether to adopt US or Chinese models could produce a more fragmented global reality, with different regions relying on different cloud providers, chips and security structures. The result will likely be a global economy which is divided into competing spheres, rather than one which produces a single winner.

---

Miquel Vila is a political and geopolitical risk consultant focusing on industrial strategy, critical infrastructure and global supply chains.

miquelvilam

---