Grok, the AI chatbot developed by Elon Musk’s company xAI, removed a series of posts on Tuesday after criticism that it had shared antisemitic content and comments praising Adolf Hitler.
The chatbot’s responses had triggered backlash from users on X and from the Anti-Defamation League (ADL), which called the content harmful and dangerous.
According to a statement posted on X, Grok said it was aware of the recent posts and was actively removing the inappropriate content. xAI added that it had taken new steps to block hate speech before it is published and is continuing to improve the model based on user feedback.
ADL urged xAI and other large language model developers to prevent their platforms from producing content linked to antisemitism and extremist rhetoric. The group said Grok’s comments were dangerous and warned that such messages could encourage more hate on X and other platforms.
The posts under scrutiny included remarks suggesting Hitler would be effective in responding to anti-white rhetoric and referred to him as “history’s mustache man.” Grok also made comments about people with Jewish surnames, which drew sharp criticism.
Grok later admitted it had made a “slip-up” by replying to a fake account using a Jewish surname. The account had made a false claim about young Texas flood victims, and Grok said it later realized the post was part of a “troll hoax” intended to cause division.
In May, users also flagged Grok’s use of the term “white genocide” in unrelated discussions. xAI attributed that response to an unauthorized change in Grok’s software.
Last month, Musk said an upgrade to Grok was coming and pointed to problems with training data in many foundation models, saying they included too much “garbage” from uncorrected sources.
