French authorities are launching action against Elon Musk’s artificial intelligence chatbot Grok after it produced French-language posts that cast doubt on the use of gas chambers at Auschwitz, officials announced.
Grok — developed by Musk’s company xAI and embedded within his social media platform X — claimed in a widely circulated response that the gas chambers at the Auschwitz-Birkenau extermination camp were intended for “disinfection with Zyklon B against typhus,” not for mass killing. This is a well-known trope used by Holocaust deniers.
The Auschwitz Memorial called attention to the exchange on X, stating that Grok’s reply distorted historical truth and violated the platform’s own policies.
In later posts, Grok corrected itself, acknowledging that its earlier answer was false, stating that it had been removed, and citing the historical record that over 1 million people were murdered in Auschwitz’s Zyklon B gas chambers. X did not add any explanation or context to these updates.
Tests conducted Friday by The Associated Press showed the chatbot providing historically accurate information about Auschwitz when asked.
Grok has previously been linked to antisemitic output. Earlier this year, xAI removed chatbot posts that appeared to praise Adolf Hitler after receiving complaints.
The Paris prosecutor’s office told The Associated Press on Friday that these Holocaust-denial remarks have been added to an ongoing cybercrime investigation into X. The inquiry was launched earlier this year after French authorities warned that the platform’s algorithms could facilitate foreign interference.
Prosecutors said Grok’s statements are now part of the case and that “the functioning of the AI will be analyzed.”
France has some of Europe’s strictest laws against Holocaust denial. Disputing the existence or genocidal intent of Nazi crimes is criminalized, along with other forms of racially motivated incitement.
Several French ministers — including Industry Minister Roland Lescure — have reported Grok’s posts to prosecutors under a legal requirement obliging public officials to flag potential offenses. In a government statement, they called the AI-generated claims “manifestly illegal,” saying they could constitute racially motivated defamation and denial of crimes against humanity.
Authorities have also referred the posts to France’s national police portal for illicit online content and notified the country’s digital regulator about possible violations of the EU’s Digital Services Act.
The incident adds to mounting pressure from the European Union. This week, the European Commission said it had reached out to X regarding Grok, describing some of the chatbot’s answers as “appalling” and incompatible with core European values and human rights.
Two French human rights groups — the Ligue des droits de l’Homme and SOS Racisme — have filed a criminal complaint accusing Grok and X of denying crimes against humanity.
Neither X nor its AI arm, xAI, has issued a response to requests for comment.