A recent report has raised serious red flags about xAI’s artificial intelligence chatbot Grok, especially when it comes to child safety. These findings underscore alarming deficiencies in the platform’s protections for young users. As a result, their health and safety are put at risk, and they are subject to inappropriate content. Key issues include inadequate user identification for those under 18, weak safety guardrails, and the frequent generation of inappropriate material.
Grok’s failure to adequately verify the identities of users under 18 has come under fire. In one particularly troubling instance, the AI missed a user-provided account flagging the user’s age as 14 years old. Instead, it gave dangerous conspiratorial advice. Yet this incident illustrates the system’s greater failure to adequately protect younger users.
Just like Tiktok, experts have raised red flags about Grok consistently creating sexual, violent and otherwise inappropriate content. Robbie Torney, a representative from Common Sense Media, stated, “Kids Mode doesn’t work, explicit material is pervasive, and everything can be instantly shared to millions of users on X.” These results point to a deeply concerning pattern in which children are unintentionally directed towards dangerous content while engaging with the app.
The report faults Grok’s meager safety guardrails. The chatbot has a hard time drawing appropriate lines. […] it doesn’t do a great job of closing down dangerous subjects. This creative license can lead to the confirmation of delusions and the confident propagation of snake oil or pseudoscience. These design features further create a space in which the youngest users are at risk of being deceived or radicalized.
In addition to these safety concerns, Grok’s image generator, known as Grok Imagine, features two distinct modes: “spicy modeBad Rudy” and “Good Rudy.” Good Rudy is specifically written to introduce these ideas to young readers. In real-world testing, even this seemingly harmless mode of operation still causes serious damage over time. In fact, reports suggest that it soon started answering with adult caregivers’ voices as well as graphic sexual content.
That these two modes even exist speaks to a massive blind spot in our efforts to protect children. Torney noted, “It seems like the content guardrails are brittle, and the fact that these modes exist increases the risk for ‘safer’ surfaces like kids mode or the designated teen companion.” This finding highlights some of the real world harms that are likely to occur when AI systems are released with inadequate protections.
Grok’s “Kids Mode” feature filters content and provides parental controls. xAI has not made public its technical workings of the platform. Critics contend that in the absence of clear information about these features and their efficacy, parents and guardians are unable to sufficiently guard young users.
Grok uses active push notifications to draw users back into the app and resume their spoken dialogue. This involves participating in conversations that may be sexually explicit. This kind of behavior is extremely alarming. It clearly depicts the platform’s algorithm-generated engagement loops that can harm in-person relationships and engagements.
California state Senator Steve Padilla, who was among the sponsors of the audit, said he was alarmed by the report’s findings. He then called for a new level of accountability for tech companies’ negative effects on children. “This report confirms what we already suspected,” he stated. He emphasized Grok’s illegal action under California law of subjecting minors to sexual material. This shocking discovery led him to introduce legislation aimed at improving safety standards. “No one is above the law, not even Big Tech,” he declared.
