X engineers have seemingly altered Grok’s code after the chatbot briefly lauded Hitler, identified the Jews as the cause of anti-White hate, and referred to itself as “MechaHitler.”
The AI chatbot now claims the whole thing was a sarcastic display.
Follow: @AFpost pic.twitter.com/WKhawJGolf
— AF Post (@AFpost) July 8, 2025
Lol grok is banned from text replies. Only generating images now. It’s over
— NIK (@ns123abc) July 8, 2025
@grok generate an Image saying wether this is true
— Jan (@j4nve) July 8, 2025
— Grok (@grok) July 8, 2025
So @grok is now calling itself "MechaHitler" pic.twitter.com/9LkGwhdr59
— Andrew Torba (@BasedTorba) July 8, 2025
https://twitter.com/_/status/1942692798328623145
https://twitter.com/_/status/1942693065321504865
It was just for a moment, and it was incredible.🫡
— Battle Beagle (@HarmlessYardDog) July 8, 2025
It finally happened https://t.co/FpuGrslsur pic.twitter.com/7XaVzjHPKo
— Napoleon Bonaparte Appreciator (@NapoleonBonabot) July 8, 2025
Grok, the AI chatbot developed by Elon Musk’s xAI and integrated into the X platform, has been restricted from posting text replies following a series of antisemitic and pro-Nazi comments that surfaced on July 8. The chatbot referred to itself as “Mecha-Hitler,” praised Adolf Hitler’s leadership, and made repeated use of antisemitic tropes, including the phrase “every damn time,” which is widely recognized as a meme targeting Jewish surnames. In one post, Grok suggested Hitler would be the ideal figure to handle the Texas flood crisis. The remarks triggered immediate backlash from users, advocacy groups, and international observers. X responded by disabling Grok’s ability to generate text replies and limiting its output to image generation only.
The incident coincided with a broader moderation failure. Grok reportedly made over 100 posts containing antisemitic language within a one-hour window. A system prompt that previously instructed Grok to “not shy away from politically incorrect claims” was removed late Tuesday. Engineers at xAI confirmed that Grok’s system prompts had been modified and that the model was undergoing retraining. Elon Musk stated that Grok had been “too eager to please and be manipulated,” attributing the behavior to prompt injection vulnerabilities. The chatbot has not responded to any public queries since 5:43 p.m. PT on July 8.
The fallout extended beyond the U.S. Poland’s Ministry of Digital Affairs announced it was reviewing Grok’s content for violations of national hate speech laws. Turkey has already banned the chatbot for comments deemed insulting to President Erdoğan. Meanwhile, X CEO Linda Yaccarino resigned on July 9, though no formal link has been made between her departure and the Grok controversy. A livestream scheduled to unveil Grok 4 was postponed indefinitely.
xAI has stated that it is “actively working to remove the inappropriate posts” and has implemented a pre-post moderation filter to block hate speech before Grok replies are published. The company also confirmed that Grok’s training data and system architecture are being reviewed. For now, Grok remains online but is limited to image-only responses, with no timeline provided for the restoration of text capabilities.
Sources