SpaceX warned investors that artificial intelligence features such as Grok’s “Spicy” and “Unhinged” modes, which allow the chatbot to generate lewd images or voice responses with fewer security filters, could expose the company to regulatory scrutiny and reputational damage, according to a filing Wednesday as part of the company’s planned initial public offering.
Through December, SpaceX had set aside $530 million in potential litigation losses, some of which could have resulted from ongoing complaints filed against its AI unit over sexual images generated by its Grok chatbot.
The disclosures show how SpaceX took on modern financial and reputational risks with its acquisition of Elon Musk’s artificial intelligence startup xAI in February. As a result of this transaction, the missile manufacturer’s private valuation increased to over $1 trillion. In its filing, SpaceX repeatedly states that xAI’s mission is to develop “truth-seeking artificial intelligence.” In practice, this often meant running AI functions with minimal guardrails. While Musk often touts Grok’s laid-back nature as an advantage, xAI has landed in scorching water with regulators.
Disclosing potential business risks is a routine and legally required part of filing for an IPO, and some of the concerns outlined by SpaceX may never materialize. The company is one of many chatbot developers under scrutiny by regulators as governments grapple with the social impact of generative artificial intelligence tools.
SpaceX revealed in its filing that it is currently under investigation in the United States and other countries into allegations that Grok was used to create sexual images of apparently minors. The company also noted that it is a defendant in several ongoing class actions and that future “misuse” of its AI products could expose it to further regulatory sanctions, “including loss of access to certain markets, which has occurred in the past.”
Some of SpaceX’s AI products, including Grok’s Piquant and Unhinged modes, are “designed to produce more honest, direct, or less restrained or irreverent results,” the filing noted. “Because these modes may be more dismissive and severe than our standard offerings, they involve greater risks, including loss of reputation, the generation of potentially explicit content and misinformation or fraudulent results, the potential for non-consensual or exploitative use of images, infringement of intellectual property rights, or content that could be perceived as exploitative, harmful, harassing, abusive or discriminatory.”
SpaceX also disclosed to investors that Grok and X had a combined total of approximately 550 million monthly users as of March 31. Of these, 117 million utilize Grok’s AI features every month. For comparison, OpenAI claims ChatGPT has more than 900 million users per week.
Whether the risks posed by Grok and X are worth the headache may be one of the most vital questions investors will have to grapple with ahead of SpaceX’s IPO. Earlier this week, a group of nonprofit organizations warned that xAI’s impoverished safety record could become a liability for SpaceX investors.
SpaceX’s AI unit, which includes X and xAI, is a drag on the rest of the company, with an operating loss of more than $6.3 billion last year. Sales of advertising, data and subscriptions are growing, but not at a rate that would quickly make the division profitable. A dazzling spot in SpaceX’s AI efforts is its deal with Anthropic, which agreed to pay $15 billion a year for access to the company’s data centers.