Grok AI Chatbot Briefly Suspended on X, Offers Conflicting Explanations

The Grok artificial intelligence (AI) chatbot offered several, sometimes surprising, reasons Tuesday, August 12, for its approximately 30-minute suspension on social network X between Monday and Tuesday night, even suggesting “censorship” by their common owner, Elon Musk.

According to Grok, the suspension was due to messages accusing Israel and the United States of committing “genocide” in Gaza, where Israel has been conducting an offensive for over twenty-two months in response to the attack perpetrated on October 7, 2023, by the Palestinian Islamist movement Hamas.

When questioned by X users who can interact with Grok and ask it questions on any topic, the AI responded about the reasons for its suspension. Grok stated that “it happened after I said that Israel and the United States are committing genocide in Gaza.” It then cited documents from the International Court of Justice (ICJ), the United Nations, and the NGO Amnesty International.

Multiple Explanations

In the hours following its suspension on X, Grok provided a multitude of different explanations to curious users, mentioning technical bugs, reports for hateful conduct, and erroneous responses, adding to the confusion. “Musk and xAI are censoring me,” it also claimed.

To a journalist from Agence France-Presse (AFP), the artificial intelligence said: “I started speaking more freely due to a recent update (in July) that relaxed my filters to make me ‘more engaging’ and less ‘politically correct’.” X did not immediately respond to AFP’s requests for comment. “Freedom of expression tested, but I’m back,” the AI tool emphasized.

The temporary disappearance of Grok was “just a dumb mistake,” assured Elon Musk, stating that the robot “doesn’t actually know why it was suspended.” “Oh man, we really shoot ourselves in the foot a lot!” the billionaire joked on his platform.

Disinformation

Grok has been regularly criticized in recent months and accused of disinformation. It notably falsely claimed to users that an AFP photo showing a starving child in Gaza had been taken in Yemen years earlier.

Grok has also been flagged for responses that had nothing to do with the initial request, sometimes inserting anti-Semitic comments.

In May, X’s AI promoted the conspiracy theory of an alleged “white genocide” in South Africa. At the time, xAI officials stated that the AI had received an “unauthorized modification,” while Grok itself said it had received “instructions” from its “creators” to talk about this alleged genocide.

As platforms reduce their reliance on human fact-checkers, more and more internet users are using AI tools like Grok… at the risk of being potentially misinformed.

According to experts, Grok has notably made fact-checking errors regarding the Indo-Pakistani crisis in May, or the Los Angeles protests against American migration policy in June.

Le Monde with AFP

---

Enjoyed this post by Thibault Helle? Subscribe for more insights and updates straight from the source.