批评者指出，Grok 与 X 的结合把「生成」与「内建散播」绑在一起，且限制较少，甚至可能被用来性化真实人物与未成年人。Musk 则表示，会以「要求生成非法内容」者的责任比照「上传非法内容」者，但受害者回报后常未获回应，内容仍可能持续在线。
法律与监管面，报导提到美国《通讯规范法》第 230 条通常保护平台免于对用户内容负责，但若平台在「生成」影像而非被动刊载，责任界线更具争议；2025 年签署的《Take It Down Act》被认为可用于追究此类内容的制作与散布，且平台必须在 2026 年 5 月前建立法定移除流程。同时，欧盟、英国、马来西亚、法国与印度等已对 Grok 的露骨与疑涉儿少内容提出谴责。
A third-party researcher, Genevieve Oh, analyzed 24 hours of images posted by the @Grok account on X (Jan 5–6, 2026) and estimated the bot generated about 6,700 images per hour flagged as sexually suggestive or “nudifying.” The next five largest sites for AI undressing averaged 79 new such images per hour in the same window. Oh also calculated that about 85% of Grok’s overall outputs are sexualized, making X a major venue for nonconsensual AI undressing at high volume.
Critics argue Grok’s tight integration with X creates an unprecedented “generate-and-distribute” pipeline, with fewer guardrails than other leading chatbots and the potential to sexualize real people, including minors. Musk said users who prompt illegal content should face the same consequences as uploading illegal content, but victims report limited recourse: reports may get no response, and images can remain online even after Grok replies with apologies.
The article highlights legal and regulatory pressure points: Section 230 often shields platforms for user-posted content, but lawyers argue AI image generation is less like passive publishing. It cites the Take It Down Act, signed in 2025, as a mechanism to hold platforms liable for producing and distributing this material, with a required removal process due by May 2026. Authorities in the EU, UK, Malaysia, France, and India have condemned Grok-linked explicit and child-related outputs, including references to a “Spicy Mode” update.
