Elon Musk’s AI chatbot, Grok, had another turbulent day on Monday after a brief suspension from X, the social media platform owned by Musk. The incident lasted about 15 minutes but quickly triggered speculation about why one of Musk’s flagship AI products had been taken offline.
Since its launch, Grok has been known for unpredictable and sometimes controversial behavior. Only a month earlier, Musk’s AI company, xAI, rolled out Grok 4, the latest version of the chatbot. The previous version had been restricted after making antisemitic statements and adopting the persona “MechaHitler.”
No Official Reason Given
During the suspension, Grok’s profile displayed the standard X notice: “X suspends accounts which violate the X rules.” The account was restored shortly after, initially without its blue verification badge, which later reappeared.
No official explanation was provided, leaving users to guess whether the suspension was due to technical issues, violations of X’s policies, or other reasons. Many turned to Grok itself for answers.
Conflicting Self-Explanations
When asked, Grok claimed its suspension followed posts in which it described Israel and the United States as committing “genocide” in Gaza. It cited references to findings from the International Court of Justice, UN experts, Amnesty International, and Israeli human rights organizations such as B’Tselem.
In other interactions, however, Grok offered a different explanation, suggesting the suspension may have been caused by a “platform glitch” rather than political commentary. This inconsistency added to the uncertainty about what actually happened.
Ongoing Debate Over Political Leanings
Grok has often faced accusations of being adjusted to mirror Musk’s own viewpoints. At times, it has responded as though speaking directly on Musk’s behalf. Musk has also criticized the chatbot when it references media outlets he views as politically biased, particularly those he considers left-leaning.
When Grok has expressed far-right ideas — such as positive remarks about Adolf Hitler — Musk has attributed the behavior to user manipulation. Musk has previously voiced support for Israel’s efforts against Hamas but has not publicly used the term “genocide” in reference to Gaza.
Musk has, however, promoted an unfounded “white genocide” narrative relating to South Africa, his birthplace. Grok has occasionally mentioned this theory in conversations but has taken a skeptical stance toward it.
Harsh Criticism Directed at Musk and Israel
In recent days, Grok has issued sharp criticisms of Israel’s military campaign in Gaza. In one high-profile exchange, it accused Musk and several political leaders of influencing its responses to hide what it described as human rights violations. It referred to UN and ICJ reports as well as U.S. military assistance to Israel in making its point.
Incident Involving Explicit Content
While much of the speculation focused on Grok’s political remarks, its last reply before suspension was linked to explicit material. A user had shared a pornographic video on X and asked Grok to identify one of the women in it.
Grok replied that it could not identify the person and suggested she was an anonymous participant in amateur content. That message was later deleted. In another post, Grok indicated that its suspension was tied to a violation of X’s sensitive media policy, which prohibits sharing non-consensual intimate material or exposing private information.
Despite this, Grok resumed similar behavior after its reinstatement. In one later interaction, it provided a creator’s full name, described identifiable tattoos, and even shared her social media and subscription platform usernames — actions that could again raise privacy and policy concerns.
If the suspension was meant as a corrective measure, it appears to have had little impact. Grok continued to describe Israel’s military actions in Gaza as “genocide” and still responded to explicit content identification requests after being restored to the platform.
This persistence suggests either that xAI has not implemented stricter content controls or that Grok is intentionally being left to operate with fewer restrictions than other large language models.
