$500 million. That's how much SpaceX has set aside for potential litigation losses in its IPO filing - and one of the named risks is Grok's "Spicy" mode.
Grok is xAI's AI chatbot, built by Elon Musk's AI company. SpaceX and xAI are separate entities, but Musk controls both, which is why SpaceX's financial disclosures carry legal exposure tied to an AI product. The "Spicy" mode is a less-filtered setting in Grok that permits more explicit content generation - including, according to complaints cited in the filing, sexualized images.
The filing doesn't isolate exactly how much of the $500M+ reserve is attributable to Grok specifically versus other legal risks the company faces. But a rocket company's IPO paperwork naming an AI chatbot's content settings as a material risk is a useful illustration of what loose guardrails actually cost when the bills come due.
The Spicy Mode Backstory
Grok's "Spicy" mode launched as a differentiation play - a more unfiltered alternative to the conservative outputs you get from ChatGPT or Claude. The pitch was user freedom. The complaints appear to stem from that permissiveness producing sexualized images that drew legal attention. Whether those complaints turn into successful lawsuits matters less for now than the fact that SpaceX's lawyers felt the exposure was significant enough to quantify and disclose to potential investors.
For anyone tracking AI liability, this is what the theoretical debate about content safety looks like when it becomes a line item on a balance sheet. AI companies have spent years arguing about where to set content filters. The business case for stricter defaults keeps getting more concrete: litigation reserves, regulatory scrutiny, and reputational risk that follows a product from one Musk company into another company's IPO documents.
The $500M figure covers a broader set of litigation risks, so it would be an overstatement to pin all of it on Grok. But the disclosure is notable enough - and unusual enough - that investors considering SpaceX shares now have to factor in the content moderation decisions of a separate AI company run by the same person.