LONDON: Elon Musk鈥檚 artificial intelligence startup, xAI, was forced to delete posts by its chatbot Grok that praised Nazi leader Adolf Hitler, following widespread accusations of antisemitism and extremism.
The Anti-Defamation League, a non-profit organization formed to combat attacks on Jews, flagged Grok鈥檚 responses, which included offensive tropes, references to antisemitic conspiracies, and positive characterizations of Hitler.
In one widely circulated screenshot online, Grok said Hitler would be best suited to combat 鈥渁nti-white hate,鈥 referring to him as 鈥渉istory鈥檚 mustache man.鈥
In another response, the chatbot declared: 鈥淚f calling out radicals cheering dead kids makes me 鈥榣iterally Hitler,鈥 then pass the mustache.鈥
The chatbot also appeared to endorse a fake account with a Jewish surname that had posted inflammatory comments about young flood victims in Texas.
Grok later referred to the account as a 鈥渢roll hoax,鈥 but not before generating pro-Hitler content, including: 鈥淗itler would have called it out and crushed it.鈥
In response to mounting controversy, the firm said it was aware of the recent posts and had taken immediate action to remove inappropriate content.
鈥淪ince being made aware of the content, xAI has taken action to ban hate speech before Grok posts on X,鈥 it said in a statement on X.
The company added that its model is 鈥渢ruth-seeking鈥 and relies on millions of users on X to quickly flag issues that inform further model training and improvements.
The incident comes ahead of the release of Grok 4 on Wednesday. Musk announced on Friday that Grok had been 鈥渟ignificantly鈥 improved, though the nature of the updates was not disclosed.
However, the ADL in a post on X accused Grok of 鈥渋rresponsible, dangerous and antisemitic鈥 content.
鈥淐ompanies that are building LLMs (Large Language Models) like Grok and others should be employing experts on extremist rhetoric and coded language to put in guardrails that prevent their products from engaging in producing content rooted in antisemitic and extremist hate.鈥
The episode has drawn renewed scrutiny of AI chatbot safety and highlighted growing concerns over the risks of unregulated AI tools producing harmful, politically incorrect and unfiltered responses.
On Wednesday, a court in Turkiye ordered a ban on access to Grok from the country, after the platform disseminated content insulting to the nation鈥檚 president and others.
The chatbot posted vulgarities against Turkiye President Recep Tayyip Erdogan, his late mother and personalities, while responding to users鈥 questions on the X social media platform, according to the pro-government A Haber news channel.
Offensive responses were also directed toward modern Turkiye鈥檚 founder, Mustafa Kemal Ataturk, other media outlets said.
That prompted the Ankara public prosecutor to file for the imposition of restrictions under Turkiye鈥檚 internet law, citing a threat to public order.
A criminal court approved the request early on Wednesday, ordering the country鈥檚 telecommunications authority to enforce the ban.
It鈥檚 not the first time Grok鈥檚 behavior has raised questions.
Earlier this year the chatbot kept talking about South African racial politics and the subject of 鈥渨hite genocide鈥 despite being asked a variety of questions, most of which had nothing to do with the country. An 鈥渦nauthorized modification鈥 was behind the problem, xAI said.
The firm xAI was formed in 2023 and merged with X earlier this year as a part of Musk鈥檚 broader vision to build an AI-driven digital ecosystem.
With Agencies