SpaceX Identifies Grok’s ‘Spicy’ Mode as a Potential Risk in IPO Submission

SpaceX has cautioned investors that features in AI, like Grok’s “Spicy” and “Unhinged” modes, which permit the chatbot to produce suggestive image or voice outputs with fewer safety protocols, might expose the company to regulatory scrutiny and reputational risks, as noted in a filing released Wednesday related to its planned initial public offering.

By December, SpaceX had allocated $530 million for potential litigation expenses, some of which could arise from ongoing complaints regarding sexualized imagery produced by its Grok chatbot.

The disclosures indicate that SpaceX embraced new financial and reputational challenges when it acquired Elon Musk’s artificial intelligence entity xAI in February, a transaction that propelled the rocket manufacturer’s private valuation past $1 trillion. In the filing, SpaceX consistently asserts that xAI aims to develop “truth-seeking artificial intelligence.” In practical terms, this has often translated to the introduction of AI features with minimal safeguards. While Musk promotes Grok’s unfiltered nature as a key advantage, it has attracted criticism from regulators.

Revealing potential business risks is a standard and legally mandated part of IPO filings, and some of the issues identified by SpaceX may never come to fruition. The company is among several chatbot developers currently under scrutiny from regulators as governments address the societal consequences of generative AI tools.

SpaceX disclosed in the filing that it is presently under investigation in the United States and other nations regarding claims that Grok was utilized to create sexualized imagery featuring apparent minors. Additionally, the company cited its involvement as a defendant in several ongoing class action lawsuits, and highlighted that future “misuse” of its AI products could lead to further regulatory sanctions, “including loss of access to certain markets, which has happened in the past.”

Some of SpaceX’s AI offerings, including Grok’s Spicy and Unhinged modes, are “designed to generate more candid, direct, or less restrained or irreverent outputs,” according to the filing. “Because these modes may be more irreverent and harsher than our standard offerings, they present heightened risks, including reputational harm, the generation of potentially explicit content and misinformation or misleading outputs, possible nonconsensual or exploitative imagery, intellectual property infringement, or content that may be viewed as exploitative, harmful, harassing, abusive, or discriminatory.”

SpaceX also informed investors that Grok and X together have roughly 550 million combined monthly users as of March 31, per the filing. Out of that total, 117 million regularly engage with Grok’s AI features each month. In contrast, OpenAI reports that ChatGPT has over 900 million weekly users.

Whether the risks associated with Grok and X are manageable will be a significant consideration for investors leading up to the SpaceX IPO. Earlier this week, a coalition of nonprofits cautioned that xAI’s poor safety history could pose a liability for investors in SpaceX.

SpaceX’s AI division, which encompasses X and xAI, is a drag on the overall company, recording an operating loss of over $6.3 billion last year. Although revenue from ads, data, and subscriptions is increasing, it is not at a pace that would swiftly lead the division to profitability. A potential positive for SpaceX’s AI initiatives is its agreement with Anthropic, which has committed to pay $15 billion annually for access to the company’s data centers.