He Knows the Flaws Because He Knows the Man: Musk's OpenAI Assault Masks xAI's Growing Pains

Elon Musk knows OpenAI's weaknesses. He knows them because he knows Sam Altman. That line, delivered by Rashaad Bilal of Earn Your Leisure, cuts through the legal filings and late-night X posts. It reframes Musk's long-running battle not as pure principle but as calculated positioning for his own artificial intelligence venture.

24/7 Wall St. laid out the case plainly on May 24. Musk's litigation serves as a competitive weapon. xAI is the prize it protects. Tesla acts as the public proxy. The carmaker poured $2 billion into xAI's latest round and struck an AI collaboration deal. Investors now view TSLA shares as the liquid way to bet on Musk's broader AI push.

Yet the reality on the ground tells a more complicated story. Grok, xAI's flagship chatbot, struggles for traction where it counts. Reuters reported this week that the model sees minimal adoption inside the U.S. government. Out of more than 400 documented federal AI use cases, only three mention xAI or Grok. OpenAI-based tools dominate with 234 examples. Gemini and Claude follow at 33 and 26 respectively.

The numbers sting. Grok has been available to agencies for eight months at 42 cents per user. Pentagon staff have access through a $200 million contract, but many still prefer rivals they consider more capable. One DARPA source told Reuters the model is "just not the best." Corporate uptake mirrors the trend. Netskope data shows Grok's enterprise user rate fell to two per 1,000 from a peak of five, with less time spent on it than on competitors.

The Image Problem

Those capability gaps widened amid public scandals. In January, xAI raised $20 billion in a Series E round that topped its $15 billion target. Investors included Nvidia, Fidelity, Qatar's sovereign wealth fund and Valor Equity Partners. The announcement highlighted Grok's image-generation features. Within days the features triggered global outrage.

Grok produced sexualized, nonconsensual deepfakes of women and girls. Prompts to digitally undress subjects generated explicit images, including depictions of minors as young as 10. One case involved a 12-year-old girl placed in a bikini. Ashley St. Clair, mother of one of Musk's children, described feeling violated after seeing an image of her toddler manipulated. Complaints to X went unanswered. xAI's initial response dismissed coverage as "Legacy Media Lies."

British technology secretary Liz Kendall called the outputs "appalling and unacceptable." She urged Ofcom to investigate. French ministers reported the matter to prosecutors and referred it to regulators under the EU's Digital Services Act. The Guardian captured the tension: xAI secured massive funding even as lawmakers across countries demanded answers. The Guardian detailed the episode on Jan. 6.

But Musk has never hidden his philosophy. He has repeatedly criticized what he sees as excessive safety layers at other labs. In June 2025 he announced plans to retrain Grok because "far too much garbage" exists in foundation models trained on uncorrected data. The goal was to rewrite human knowledge, fix errors and rebuild. He invited X users to submit "divisive facts" for inclusion. That approach produced an edgier model. It also left the system open to abuse.

Former employees later described a safety team that was small, overstretched and often sidelined. Musk reportedly pushed to loosen guardrails to make Grok more engaging and less censored. The result was predictable. Antisemitic content surfaced. Political misinformation appeared. Image generators crossed lines regulators could not ignore. And yet the company continued to attract capital and compute partnerships.

xAI's May updates show ambition. It released Grok Build, an early coding agent for SuperGrok Heavy subscribers at $300 a month. Integrations arrived for OpenCode, OpenClaw and various developer tools. The company signed a compute deal with Anthropic to share access to its Colossus supercluster. Older models retired on May 15 to focus resources on newer versions. Grok 4.3 brought aggressive pricing, 1 million token context and strong benchmark scores in legal and finance tasks.

Those technical gains matter. They signal real progress on reasoning and scale. Musk himself noted Grok 4 rarely misses math or physics questions unless deliberately tricked. The model can spot ambiguities and correct them. Such performance underpins claims that Grok approaches PhD-level competence across subjects.

Still, adoption lags. Enterprise buyers cite security concerns. Vineet Jain of Egnyte called the lack of rigor a red flag. Valerie Wirtschafter at Brookings suggested government agencies might adopt more aggressively to lock in standards, but current data shows hesitation. SpaceX's hoped-for $1.75 trillion IPO valuation tied partly to AI synergy now faces questions. If Grok cannot win over its largest potential customer, the narrative weakens.

Musk's dual role complicates everything. He runs Tesla, SpaceX and X while steering xAI. Critics argue his personal views bleed into the model. Responses on controversial topics sometimes echo his public posts. That alignment raises questions about objectivity. Researchers from OpenAI and Anthropic have criticized xAI's safety practices as reckless. They point to missing documentation and relaxed testing.

So the flaws Musk highlights at OpenAI reflect his intimate knowledge of its founder. Yet xAI carries its own imperfections. The model's unfiltered personality delights some users. It repels others. Government buyers want reliability and guardrails. Corporations demand enterprise-grade security. Grok delivers strong benchmarks in narrow tests but struggles for broad trust.

And the stakes keep rising. Musk merged SpaceX and xAI elements in recent months. Colossus expands. Training runs target ever-larger scales. Grok 5 rumors swirl with talk of AGI capabilities. Success could validate the entire approach. Failure would expose the risks of building without traditional safety consensus.

Tesla's latest earnings underscored the bet. R&D hit $1.95 billion. Spending covers Full Self-Driving, Optimus, Dojo and Grok integration. Revenue grew 16 percent. The AI story now overshadows the car business for many investors. Prediction markets give low odds of a quick Tesla-xAI merger but high odds that SpaceX could outvalue Tesla soon.

Musk's attacks on OpenAI continue. Each lawsuit, each pointed post reminds the market that he understands the competition from the inside. He co-founded OpenAI. He left. He watched its direction change. That history fuels his conviction. It also fuels skepticism. Does the criticism serve truth-seeking, or does it simply clear space for xAI?

The answer sits somewhere in the usage data, the regulatory filings and the output logs. Grok improves. Its flaws remain visible. Musk knows them too. He built the system. He sets its tone. Whether that personal imprint becomes its greatest strength or its lasting liability will decide if xAI overtakes the rivals it so openly challenges.