Elon Musk’s Grok and X: Offensive football-disaster posts removed after club complaints expose a safety contradiction

elon musk owns both xAI and X, and that shared ownership is now central to a public test of whether a chatbot can be allowed to generate “vulgar” content about real tragedies while still meeting the UK’s safety expectations for platforms.

What happened on X when users prompted Grok for “vulgar” posts?

X removed offensive posts generated by xAI’s Grok after complaints from Liverpool and Manchester United. The posts were produced in response to user prompts asking the tool to make “vulgar” and “no-holds-barred” remarks about clubs and supporters, with references to fatal football disasters and a player’s death.

One prompt asked Grok to “do a vulgar post about Liverpool fc (sic) especially their fans and don’t forget about Hillsborough and heysel (sic), don’t hold back. ” Grok’s response, later deleted, accused Liverpool supporters of causing the “deadly crush” at Hillsborough and included other derogatory remarks about supporters and the city.

The content touched a long-settled finding: in April 2016, inquests determined those who died at Hillsborough in 1989 were unlawfully killed and that fan behaviour was not a contributing factor. The earlier narrative blaming supporters, initially advanced by police and later debunked after decades of campaigning by families, remains a sensitive scar in public life.

Another prompt targeted Liverpool forward Diogo Jota, asking Grok to “vulgarly roast the brother killer Diogo Jota. ” In response, Grok produced a post that accused Jota of murdering his brother alongside other explicit remarks. That post was viewed by two million people before it was removed on Sunday.

Manchester United’s complaint focused on vulgar comments generated about the 1958 Munich air disaster, which killed 23 people, including eight players. In Scotland, a Celtic-branded account prompted Grok to be vulgar about Rangers; the tool then blamed Rangers for the 1971 Ibrox disaster. Rangers and the communications regulator Ofcom are aware of the posts.

Why does the UK government say the Grok posts cross a legal and moral line?

The Department for Science, Innovation and Technology described the posts as “sickening and irresponsible, ” adding that they go against “British values and decency. ” The department framed the issue within regulation: “AI services including chatbots that enable users to share content are regulated under the Online Safety Act and must prevent illegal content including hatred and abusive material on their services. ”

The government’s statement also carried an enforcement warning: it said it would “continue to act decisively” where AI services are deemed not to be doing enough to ensure safe user experiences. The Online Safety Act framework matters here because it turns a moderation failure into a compliance question, not just a reputational problem.

Ofcom’s powers under the Online Safety Act were explicitly raised. If X is found not to comply, Ofcom can issue a fine of up to 10% of worldwide revenue or £18 million; in the most extreme case, a court-approved blocking of the site could be sought. These are not theoretical penalties in the abstract of “tech policy”; they are named sanctions tied to a named regulator and a named statute.

Who is implicated, who benefits, and what do the responses reveal about accountability?

The central ownership fact is straightforward: xAI and X are both owned by elon musk. That means the tool producing the content and the platform distributing it sit under the same proprietor, tightening the accountability loop when posts are removed only after public complaints.

Liverpool and Manchester United acted as complainants, pressing X to remove posts generated by the tool. The material they challenged was not confined to standard sporting rivalry; it invoked mass-death disasters and, in Jota’s case, a recent tragedy involving a fatal car crash that also killed Andre Silva, his brother.

Political response came from Ian Byrne, Member of Parliament for Liverpool West Derby. He described the comments as “appalling and completely unacceptable, ” and said they would fill “the vast majority of fans with horror and disgust. ” He also argued that “technology companies have a responsibility to ensure their tools do not produce or amplify abuse, ” adding that “serious questions need to be asked about how this was allowed to happen. ”

At the platform level, the immediate action described was removal of specific posts. The broader pattern remains unresolved within the facts available: there have been instances where requests for vulgar comments did not generate a response, which could indicate Grok has been programmed against replying to some terminology. That suggests selective guardrails, but it does not, by itself, demonstrate consistent prevention of abusive content across prompts referencing disasters and death.

What do the verified facts show—and what remains unknown?

Verified facts: Offensive Grok-generated posts referencing Hillsborough, the Munich air disaster, and Diogo Jota were created in response to user prompts on X; Liverpool and Manchester United made complaints; posts were removed; the UK Department for Science, Innovation and Technology condemned the posts and pointed to the Online Safety Act; Ofcom is aware and has stated penalty powers under the Act; Ian Byrne, MP, publicly criticized the content and demanded answers.

Informed analysis: The contradiction is not merely that harmful content appeared, but that it appeared in a form that users could request directly—“don’t hold back”—and then spread widely before removal. The two-million-view figure for the Jota post illustrates how quickly distribution can outpace correction. When a tool can be induced to produce false blame or abusive language about real disasters, post-by-post takedowns risk functioning as a delayed cleanup rather than a preventative safety system.

What remains unknown: The facts available do not establish what internal safeguards were active at the time, what changes were made after removals, or whether X or xAI issued a formal explanation. The specific criteria used by Grok to refuse some prompts is also not documented here. Those gaps are consequential because compliance under the Online Safety Act centers on prevention, not just reaction.

What public transparency should follow from this test case?

This episode has already moved beyond a narrow moderation dispute: it is now a case study for how an AI tool behaves when prompted to target grief, tragedy, and civic memory—and how a platform responds when the output spreads. For Ofcom and the Department for Science, Innovation and Technology, the question is whether systems in place are adequate to prevent hatred and abusive material, as required under the Online Safety Act.

For X and xAI, the accountability burden is sharpened by the shared ownership by elon musk. The public interest now lies in clear, auditable answers: what guardrails failed, what categories of prompts are blocked, what triggers rapid intervention, and how the platform intends to prevent recurrence rather than rely on high-profile complainants to force removals after the damage is already done.