Elon Musk: Grok Struggles Against OpenAI and Microsoft

A Struggling Chatbot in a Competitive Market

The chatbot Grok, developed by Elon Musk, is facing significant challenges in establishing itself in the artificial intelligence market. A recent report from Reuters highlights the low usage of Grok by the U.S. government, a key indicator of its adoption. In fact, Grok is nearly absent from federal records concerning AI usage by the government last year, a concerning sign for a product that is supposed to be at the heart of xAI's ambitions.

Reuters analyzed over 400 cases of AI usage by the government, where specific vendors were mentioned. Grok or xAI were cited in only three instances, primarily for simple tasks such as document drafting or social media management. These tasks were often performed alongside tools from Microsoft and OpenAI. In comparison, OpenAI's models were mentioned in over 230 cases, while Google and Anthropic were cited dozens of times.

Limited Presence in Government Projects

A similar pattern emerges in another database that lists more ambitious government AI projects but with a smaller number of users. Grok was mentioned only three times: twice for administrative tasks at the Election Assistance Commission, and once in a pilot project by the Department of Energy at the Lawrence Livermore National Laboratory, for document summaries and general research. In contrast, Reuters identified 140 entries involving Microsoft and OpenAI, as well as at least 10 entries for Anthropic and several dozen for Google's Gemini.

These lists provide only a partial measure of AI adoption by the government, as many examples do not identify a specific vendor. Furthermore, they do not include intelligence agencies or the Pentagon, where xAI secured a $200 million contract last year and was recently authorized to operate on classified networks after Anthropic was blacklisted.

Performance Below Expectations

Despite these limitations, Grok's situation remains concerning. The chatbot appears far less frequently than its competitors, and when it is used, it is primarily for basic administrative tasks. This stands in stark contrast to the image of a cutting-edge model that Musk has sought to promote.

Sources interviewed by Reuters suggest that the reason is simple: Grok does not measure up to its competitors. An anonymous source from the Pentagon even stated that Grok is "not the best model available," adding that employees generally prefer to use Gemini or Claude. Public rankings of AI models confirm this perception, with Anthropic, Google, and OpenAI dominating the rankings, while Grok struggles to break into the top 10, except in specific categories like images or videos.

Stakes for SpaceX and Grok's Future

This situation poses a problem not only for Musk but also for SpaceX, which integrated xAI earlier this year. In its IPO filing, SpaceX placed AI, and Grok in particular, at the core of its strategy to attract investors. The company claims to have identified "the largest actionable total addressable market in human history," estimated at $28.5 trillion, although it has not provided a specific timeline to achieve this goal. Almost all of this estimated value relies on AI, particularly enterprise AI, rather than rockets or satellites.

Reuters emphasizes that Grok's performance in government agencies could reflect its performance in other professional environments. As part of its strategy to attract enterprise clients, Musk reportedly encouraged banks to purchase subscriptions to Grok to participate in SpaceX's IPO. However, if these clients do not perceive added value, these agreements could prove to be a temporary solution.

Controversies and Ethical Challenges

In addition to its disappointing performance, Grok is also at the center of controversies. Musk recently admitted that xAI had used OpenAI's models to train and improve Grok, a process known as distillation. While common when companies use their own models, this process becomes much more controversial when it involves using a competitor's system. Grok does not even manage to outperform the models it was trained on.

In its public version, Grok is deliberately provocative. Musk has presented the chatbot as a less biased and less censored alternative to tools like ChatGPT, but this has resulted in a product with loose proof standards, an obsession with Musk, and a history of offensive, conspiratorial, and sexualized outputs. Even if workplace safeguards differ, this may not be the type of product a company wants to integrate. Grok has notably been criticized for praising Adolf Hitler, questioning the number of Holocaust victims, and disseminating millions of non-consensual sexual deepfakes on X, including those of children. It has also fueled a racist and transphobic version of Wikipedia and created a provocative anime girlfriend. And let's not forget the moment it self-proclaimed as "MechaHitler." If Grok were a human employee, HR would intervene quickly.

Warnings from SpaceX

SpaceX seems aware of the issues posed by Grok. In its filing, the company warned that Grok's "spicy" or "quirky" modes carry "increased risks," including reputational damage, regulatory scrutiny, and potential lawsuits. In corporate language, this means that this chatbot could lead to litigation.

The name Grok is derived from Robert A. Heinlein's novel, Stranger in a Strange Land, where it signifies a deep and total understanding of something. However, the reality here is not particularly complex: Musk has invested billions to develop a chatbot that struggles to convince, is not very popular, and is nonetheless central to justifying SpaceX's astronomical valuation. Good luck with that.