Switzerland's finance minister just asked a prosecutor to answer a question that could reshape the internet: when an AI insults a public figure, who goes to court—the human who prompted it, or the machine that generated it?
Karin Keller-Sutter filed a criminal complaint last month against an X user who requested Grok, the chatbot owned by Elon Musk's xAI, to "roast" her. The complaint targets the user for defamation and verbal abuse—but it also puts X itself in the crosshairs by asking prosecutors to examine whether the platform bears responsibility for failing to block Grok's vulgar, misogynistic output.
The finance ministry described the Grok response as "blatant denigration of a woman" and emphasized that "such misogyny must not be seen as normal or acceptable." That framing reveals the real target: not just one user's cruelty, but the entire infrastructure that makes AI-generated harassment trivially easy to produce.
The tension at the heart of this case has no clear winner yet. On one side, Keller-Sutter argues that humans cannot hide behind AI as a shield against accountability. If a person deliberately prompts a chatbot to degrade someone, they own that output—regardless of whether they typed the words themselves. On the other side, Musk has actively promoted Grok's roasts as entertainment, framing the chatbot's vulgarity as a feature rather than a bug. Under that logic, the user simply asked for a joke; the AI delivered it.
But the complaint goes further, targeting X itself. Swiss law criminalizes defamation and verbal abuse, with penalties that can include fines and imprisonment. If prosecutors accept the argument that a platform can be liable for hosting AI tools that generate harassing content, X faces exposure not just in Switzerland but globally. Every jurisdiction where Grok operates would have to grapple with whether Musk's "free speech" vision is compatible with local laws against hate speech and defamation.
This case lands at the intersection of two unresolved questions. First: can a human dodge liability by claiming an AI acted autonomously? Courts have not settled whether prompting an AI to produce harmful content constitutes the same intent as producing it directly. Second: what duty do platforms have to prevent their AI systems from generating targeted harassment? X currently treats Grok's edginess as a competitive differentiator—but that positioning becomes costly if it invites criminal prosecution.
The stakes extend well beyond one Swiss official's dignity. A precedent holding X liable could force fundamental changes in how AI platforms handle public figures, potentially requiring content filters that many companies have resisted. A precedent letting the platform off the hook signals that AI-generated harassment operates in a legal gray zone, open season for anyone with an account and a grudge.
Keller-Sutter is not the first public figure targeted by AI insults, and she won't be the last. But she may be the first to fight back with criminal law rather than civil suits or platform complaints. Whether her approach succeeds or not, it signals that governments are done treating AI harassment as a technical glitch rather than a legal violation. The age of using AI as a buffer between harassers and accountability is ending.