Elon Musk’s artificial intelligence chatbot Grok has found itself in deep trouble in France, as it generated posts denying the use of gas chambers at Auschwitz for mass murder, which was enough to start a formal investigation by French authorities.
The outcry began with a response by Grok, an AI chatbot Musk’s xAI created for his social media platform X, to a user query in French that stated that gas chambers at the Auschwitz-Birkenau death camp were intended for “disinfection with Zyklon B against typhus” rather than to kill people. Such expressions are echoes of rhetoric espoused by Holocaust deniers and square directly against settled historical fact.
The Auschwitz Memorial promptly flagged the exchange via X, citing distortion of historical reality and a direct violation of the platform’s own rules. The incident has triggered widespread concern about the reliability and safety of AI-generated content, particularly on sensitive historical topics.
Musk’s Grok AI Under Scrutiny in France After Holocaust Denial Post
Grok later issued corrections, acknowledging its mistake, deleting the original response, and pointing to historical evidence that gas chambers at Auschwitz were, in fact, used to murder more than 1 million people with Zyklon B. But those follow-up messages came without any official clarification or statement from X itself.
When The Associated Press tested the chatbot on Friday, it seemed to provide historically accurate information about Auschwitz.
The Paris prosecutor’s office confirmed the addition of Grok’s Holocaust-denial comments to an existing cybercrime investigation into X, first opened earlier this year after French officials complained that the platform’s algorithm could be manipulated for foreign interference.
Prosecutors have said that “the functioning of the AI will be examined” in their inquiry. The scrutiny of the functioning is very timely, with regulators worldwide grappling with how to oversee increasingly powerful AI systems.
France takes Holocaust denial very seriously. It has some of the toughest Holocaust denial laws in Europe; contesting the reality or genocidal nature of Nazi crimes is prosecutable as a criminal offense, along with other forms of incitement to racial hatred.
A number of French government ministers, including Industry Minister Roland Lescure, have already taken the step to formally report Grok’s posts to prosecutors in accordance with a legal provision requiring public officials to flag possible crimes. In a joint government statement, they called the AI-generated content “manifestly illicit” and said it could possibly amount to racially motivated defamation and denial of crimes against humanity.
This isn’t the first time Grok has generated antisemitic content: earlier this year, Musk’s xAI company had to take down posts from the chatbot that appeared to praise Adolf Hitler following complaints about antisemitic material. The recurring issues raise questions about the safeguards built into the AI system.
French Authorities and Civil Rights Groups File Complaints Over “Appalling” AI Content
French authorities have forwarded the suspect posts to a national police platform assigned to monitor unlawful online content. They’ve also informed France’s digital regulator of suspected infractions of the European Union’s Digital Services Act, which imposes strict rules for online platforms operating in the bloc.
The case is adding to mounting pressure from Brussels. This week, the European Commission said it’s in contact with X regarding Grok and called some of the chatbot’s output “appalling” and running counter to Europe’s fundamental rights and values.
Two of Frances most prominent civil rights organizations, the Ligue des droits de l’Homme and SOS Racisme, have filed their own criminal complaint against both Grok and X – accusing them of committing crimes against humanity.
Neither X nor xAI immediately commented on the investigation or the steps being taken to prevent similar incidents in the future.
The case, however, has exposed the increasing difficulty by governments in policing AI-generated content, especially when it pertains to historical facts and sensitive topics. In a world where AI chatbots are being used increasingly, it serves as a grim reminder of the potential real-world consequences when these systems spew out harmful or incorrect information.
