Swiss government minister files digital abuse complaint against AI chatbot Grok

Swiss Finance Minister Karin Keller-Sutter is taking action against a post made by an X user and created by Grok. Keystone / Til Buergy

Switzerland's finance minister, Karin Keller-Sutter, has lodged a formal complaint over online abuse linked to the Grok AI chatbot. It is a bold signal against digital abuse and marks a clear break from the US understanding of free speech.

Elon Musk’s artificial intelligence model Grok has courted controversy in recent months. The model, which at one point called itself “MechaHitlerExternal link”, has been found to readily generate insulting posts. For a period in early 2026, it also created thousands of sexualised deepfake imagesExternal link of women and children per hour, prompting the US magazine The NationExternal link to dub it “pedo chatbot”. The European Union has since launched a wide-reaching investigationExternal link into Grok over the deepfakes.

Now Switzerland’s finance minister is taking legal action against the model over written sexualised vilification. On March 10, a Swiss pensioner asked Grok in vulgar and sexualised terms to “roast” Keller-Sutter. He then shared the AI-generated post on X (former Twitter), which has since been deleted.

Women most affected by digital abuse

Women are particularly affected by online abuse. Simone Eymann of the Public Discourse Foundation explained that the consequences of this abuse, ranging from sleep disorders to anxiety attacks, are well documented. “At the same time, victims of digital abuse sometimes withdraw from public life, which has clear implications for our democracy,” she said.

Politically engaged women are more exposed to online public abuse than men. In Germany, nearly one in four women affected by this has considered ending their professional engagement as a result, a study by the Technical University of Munich and HateAidExternal link has found.

Eymann welcomed Keller-Sutter’s legal action. “When public figures speak up rather than remain silent, they send an important signal – that digital abuse must be taken seriously,” she said.

With her complaint, the minister is sending a clear signal against digital abuse. At the same time, when a member of government takes legal action over insults, it also raises questions about freedom of speech and expression.

What kind of statements must politicians tolerate?

Jordi Calvet-Bademunt is a senior research fellow at The Future of Free SpeechExternal link, a think tank at Vanderbilt University in the United States that studies freedom of expression. In the US, Calvet-Bademunt explained, “offensive speech directed at politicians receives greater legal protection”.

“Whether the specific content at issue in this case qualifies for protection under Swiss law will depend on the particulars of the message,” he said.

The think tank advocates a broad interpretation of freedom of expression. Calvet-Bademunt said free speech “does not only protect speech that is inoffensive,” but also “offensive, shocking or disturbing” statements. This protection is particularly important for political speech, he added, “including speech about politicians”.

“This does not mean that society must endorse or condone every form of protected speech, but shielding controversial expression is fundamental to the health of our democracies,” he said.

The researcher views Keller-Sutter’s legal complaint as “part of a broader trend in democracies” where scepticism towards AI is leading politicians to call for tighter regulations.

“India confronted Google when its AI model suggested that some experts considered Prime Minister Modi’s policies to be ‘fascist’,” he said. “Turkey partially blocked access to Grok after the AI chatbot generated responses insulting President Erdogan, Atatürk, and religious values.”

Turkey is not considered a liberal democracy by international standards. Calvet-Bademunt also points to Poland, an EU member state where Prime Minister Donald Tusk has pushed back against posts generated by GrokExternal link.

Vance’s speech in Munich on free speech in Europe

What makes X stand out is that apart from numerous bots, authorities and politicians are still using the platform. When Keller-Sutter held the rotating Swiss presidency in 2025, she also communicated via X.

As president, she caused a stir when she hailed JD Vance’s speech at the Munich Security Conference in early 2025 as a “plea for direct democracy”.

In his speech, Vance used several examples to argue that, in his view, free speech in Europe was faring poorly. Among these examples were police operations in Germany “against citizens” accused of posting “anti-feminist comments online”. Shortly afterwards, Keller-Sutter clarified her praise of Vance’s speechExternal link.

SWI swissinfo.ch

Earlier this month, a ministry spokesperson told the German-language newspaper NZZ am SonntagExternal link that she had only praised part of Vance’s speech. Vulgar insults, the spokesperson said, did not fall under freedom of expression and “are a punishable crime in Switzerland”. He added that hate against women must be combated.

Is Switzerland facing a conflict with the US?

The NZZ am Sonntag headline suggests that Keller-Sutter’s legal action also has a geopolitical dimension: “An open conflict with the US is now on the horizon,” the paper stated. The complaint was filed against unknown persons, her spokesperson told Swissinfo, adding that it was now up to the public prosecutor’s office to decide whether to pursue the pensioner who wrote the prompt, or others.

Swiss authorities must also determine what responsibility Grok’s operator, Elon Musk’s X, bears. X Switzerland GmbH has so far not commented on the legal action, while a request by Swissinfo also went unanswered.

US President Donald Trump’s administration often pushes back hard against European regulation of US tech companies. It has even gone as far as imposing sanctions on officials who play a key role in regulating AI and social media in the EU. The Swiss government could therefore face a conflict.

Free speech in Trump’s America

Calvet-Bademunt also views the US administration as part of the wider negative global trend in freedom of expression. One example he cites is the “sustained pressure from President Trump’s administration” on the AI operator Anthropic, whose chatbot is seen as too “woke”.

The US has long taken a different approach to free speech than Europe, but this now seems to be crumbling among the public.

Only 61% in the US say that insulting the flag should be allowed

According to the Future of Free Speech Index 2025, support among Americans for unrestricted free speech is waning. In 2024, around 60% of them were of the opinion that freedom of expression should cover statements insulting minorities, which is about five percentage points lower than in 2021. At the same time, 61% think that freedom of expression should include insults to the US flag, down 10 percentage points since 2021.

The index also shows that in the US, most people distinguish between human insults and AI-generated insults. Only a minority believe that AI models should be able to generate content that offends their religion or the US flag. One in five support AI deepfakes of politicians.

Keller-Sutter’s move against Grok is likely to find support in Switzerland. A recent survey by the gfs.bern research institute shows that large tech firms have a poor reputation. Nine in 10 people in Switzerland see them as purely profit-driven, while 84% are concerned about “too much political influence from their countries of origin”.

Only a few distribute hate speech online

Only a few people distribute hate speech online, according to a study by the federal technology institute ETH Zurich and the Public Discourse Foundation published last December.External link

“Only 5% of users write 78% of hate speech. Targeted moderation can make a big difference,” said Eymann of the Public Discourse Foundation.

However, major platforms like X, Facebook or Instagram have significantly scaled back content moderation. Eymann therefore appeals to users: “Empathetic counter-speech works. It affects not only the authors of harmful content, but also everyone who reads the response.”

She deems it possible that “generative AI tools like Grok” will contribute to further tightening the “scale and speed of hate speech” online. Offensive or degrading content can now be quickly generated at the click of a button. “What used to take time and effort can now be created within seconds, and this also applies to hate speech,” Eymann said.

Edited by Reto Gysi von Wartburg. Adapted from German by Billi Bierling/gw