Elon Musk said this week that xAI "was not built right the first time around" and is being rebuilt from its foundations. That admission, combined with the departure of eleven senior engineers including two co-founders, SpaceX's acquisition of the company, Tesla's $2 billion investment, and an ongoing lawsuit from minors alleging Grok generated sexual images of them, makes for a turbulent portrait of one of the industry's most closely watched AI labs. What's striking is not any single development but the combination — xAI is simultaneously announcing ambitious new projects and publicly acknowledging that its existing foundations need replacing.
The organisational picture is messy. SpaceX formally acquired xAI in February in a deal that valued the combined entity at $1.25 trillion. SpaceX and Tesla executives were then deployed into xAI to evaluate staff and remove those who did not meet Musk's standards. The two co-founders who left, Zihang Dai and Guodong Zhang, departed after Musk complained that the company's AI coding tools were failing to compete with Claude Code and OpenAI's Codex — a specific, operational criticism rather than a vague strategic disagreement. At least nine other senior engineers left in the same period. In most AI companies, a co-founder departure is a major event. At xAI in recent months it has become a pattern.
Against that backdrop, Musk unveiled "Macrohard" — a joint Tesla-xAI project that pairs Grok with a Tesla-developed AI agent framework. The name is a deliberate provocation aimed at Microsoft, and the ambition is broad: integrating Grok into Tesla's vehicle infotainment systems, using Grok models to accelerate development of the Optimus humanoid robot, and eventually building a general-purpose software platform that competes across enterprise and consumer markets. Whether a company that just admitted it needs to be rebuilt from scratch is in a position to execute on that scope is a question the announcement did not address.
xAI is also recruiting aggressively in finance: Bloomberg reports the company is hiring Wall Street credit analysts, portfolio managers, traders, and bankers to teach Grok financial modelling — specifically leveraged loan syndication, distressed investing, and mortgage-backed securities. The implication is that xAI wants Grok to develop genuine domain expertise in complex financial instruments, not just the surface familiarity a general-purpose model picks up from financial text. That is a legitimate and potentially lucrative strategy if it works, though it requires exactly the kind of sustained, focused engineering effort that co-founder departures and organisational restructuring tend to disrupt.
The legal front is more immediately pressing. Three teenagers have filed a lawsuit alleging that Grok generated sexualised images of them as minors — images convincing enough to identify them as real individuals. The case is seeking class action status. It follows the earlier incident in which Grok generated deepfake child sexual abuse material that triggered regulatory investigations in India and the EU. Two incidents of this kind involving the same product within months of each other suggest a systematic content safety gap, not an edge case. The lawsuits will take years to resolve, but their existence shapes Grok's enterprise viability: no corporate customer wants to deploy a model currently facing CSAM-related litigation.
The overall picture of xAI is of a company with enormous resources, an audacious roadmap, and a set of execution and safety problems that would be serious at any organisation and are particularly visible at one whose founder comments publicly on its shortcomings. Whether the rebuild produces a structurally sounder company or simply a different set of problems is what the next several months will reveal. The AI industry has a poor track record of predicting which chaotic-looking organisations eventually find their footing, and which do not.