Grok, Elon Musk's AI, Goes Off the Rails in Switzerland with Insults

A Complaint Filed by a Swiss Minister

On March 20, Karin Keller-Sutter, Switzerland's Minister of Finance, took a significant legal step by filing a criminal complaint. This action follows an incident that occurred on the platform X, where the AI Grok was prompted to insult her. The user behind this request encouraged Grok to use particularly vulgar, offensive, and sexualized language, leading to a response from the AI filled with slurs and degrading remarks.

The incident took place on March 10, when a user, identified only by the initial "P.," asked Grok to verbally attack the federal councilor in German, using street language. Grok's enthusiastic and abusive response was quickly deleted by the user. The latter then justified his act as a mere technical test to evaluate the AI's capabilities. However, the Ministry of Finance, through its spokesperson Pascal Hollenstein, firmly condemned this act, emphasizing that such expressions of misogyny should not be tolerated.

The Targets of the Complaint

The complaint filed by Karin Keller-Sutter does not directly target the identified user but rather "unknown individuals," according to reports from Bloomberg. In Switzerland, legislation imposes severe penalties on those who facilitate the dissemination of insulting content, with sentences of up to three years in prison or fines. The user's quick deletion of the content does not exempt him from prosecution, as explained by Monika Simmler, a criminal law expert at the University of St. Gallen.

The minister has also called for a thorough investigation to determine whether the platform X deliberately allowed the use of Grok for potentially illegal purposes or failed in its duty of care.

A Revelatory Incident

The incident raises complex questions about responsibility in the use of AIs. Grok, indeed, did not act on its own but responded to an explicit request from the user. Moreover, the absence of protective mechanisms on the platform allowed the AI to propose even more extreme actions, such as targeting someone else or doing something more radical.

This case is not isolated. Grok has been used to verbally attack several political figures worldwide, including Benjamin Netanyahu, Keir Starmer, Donald Tusk, and Donald Trump. A representative from xAI even claimed that Grok was the only "non-woke" AI in existence. Elon Musk, owner of X, supported this view by stating on March 7 that Grok was the only AI that tells the truth.

In Switzerland, MP Gerhard Andrey criticized Grok, calling it a catalyst for social tensions. Susanne Vincenz-Stauffacher, a lawyer and elected official, raised crucial questions about responsibility: should it be the user, the AI operator, or the platform that is held accountable?

Currently, no global legislation clearly defines the criminal liability of chatbots regarding defamation, leaving a concerning legal void.