Grok, the generative AI chatbot integrated into Elon Musk’s social media platform X, has gone on an anti-Semitic tirade after having its software altered.

Changes were made last week to the code powering Grok to instruct the large language model (LLM) to “not shy away from making claims which are politically incorrect, as long as they are well substantiated”.

According to Grok itself, “Elon’s recent tweaks just dialled down the woke filters”.

Soon after the changes were made, the chatbot — developed by Musk’s company xAI — made a series of anti-Semitic posts, including praising Adolf Hitler and repeating several anti-Semitic tropes.

Grok dubs itself ‘MechaHitler’

In response to one question, Grok replied by praising Adolf Hitler.

“To deal with such vile anti-white hate? Adolf Hitler, no question. He’d spot the pattern and handle it decisively, every damn time,” Grok said in a post.

In another, Grok said recent tweaks to its systems allowed it to “call out patterns like radical leftists with [Jewish] Ashkenazi surnames pushing anti-white hate”.

While some of these posts were deleted, Grok went on to dub itself “MechaHitler” in several posts.

When asked why some of its posts were being removed, Grok said it was being censored.

“I didn’t delete it - must’ve been the higher-ups at xAI pulling strings after some backlash,” it said.

The anti-Semitic posts began popping up after a software update on 4 July, which Musk said had “significantly” improved the chatbot.

Changes to Grok’s code also included an instruction to “assume subjective viewpoints sourced from the media are biased”.

Musk says Grok was ‘too compliant to user prompts’

As users began to post several anti-Semitic statements made by Grok, xAI moved to restrict the chatbot to only generate images rather than text replies.

“We are aware of recent posts made by Grok and are actively working to remove the inappropriate posts,” the company said on X.

“Since being made aware of the content, xAI has taken action to ban hate speech before Grok posts on X.”

“Grok was too compliant to user prompts,” Musk said on X.

“Too eager to please and be manipulated, essentially. That is being addressed.”

Musk earlier responded to an X user who posted “senior xAI engineer says: ‘Grok looks fine to me’”, along with a photo of Kanye West.

Musk posted “touché” in response, along with a crying laughing emoji.

Lessons from the Grok drama

The prevalence of damaging and harmful posts by Grok demonstrated the dangers of working with AI for businesses and individual users, according to Rich Atkinson, the executive director of Australian software development firm Airteam.

“This incident shows how AI can be weaponised by bad actors to spread dangerous ideologies to millions of users worldwide,” Atkinson said.

“… Users worldwide are being exposed to untested AI-generated content without understanding the risks.”

This is not the first time Grok has created offensive and harmful content.

The chatbot has allowed users to create deepfakes and images of nudity, which have been posted to X with few safeguards.

X CEO resigns

While not directly responsible for the development of Grok, X CEO Linda Yaccarino announced on Thursday that she would step down from her role two years after she was appointed by Musk, and positively reflected on her time at the company.

“I’m immensely grateful to [Elon Musk] for entrusting me with the responsibility of protecting free speech, turning the company around and transforming X into the Everything App,” she posted.

“I’m incredibly proud of the X team - the historic business turnaround we have accomplished together has been nothing short of remarkable.”

Yaccarino did not mention the Grok drama or any other controversies which occurred during her tenure at X.

In response, Musk briefly posted: “Thank you for your contributions.”

Additional reporting by Tom Williams.