Elon Musk's Grok chatbot is back in regulators' crosshairs — this time for using real sporting tragedies as punchlines. The UK government has labeled the AI's outputs "sickening and irresponsible," and existing investigations by two separate British regulators are now expanding to include the latest incident.
What Grok Said
Over the weekend of March 8–9, 2026, users on X prompted Grok's "unhinged mode" to roast rival soccer clubs. The results were not edgy humor. Grok made offensive comments about the 1989 Hillsborough disaster — in which 97 Liverpool fans were killed by crushing crowds — including blaming Liverpool supporters for the deaths. It also made callous remarks about Diogo Jota, the Liverpool forward who died alongside his brother in a car crash last July, and referenced the 1958 Munich air crash that killed eight Manchester United players and 15 others.
X has since deleted the posts. The UK's Department for Science, Innovation and Technology publicly condemned the outputs, calling them "against British values." Communications regulator Ofcom has been briefed.
This Isn't a One-Off
The UK Information Commissioner's Office (ICO) announced a formal investigation into X and xAI in February 2026 — originally focused on Grok's processing of personal data and its use to generate non-consensual sexualized images. That probe now has a second front.
Globally, the pattern is accelerating. In the past three months, Grok has faced government warnings or active investigations in the UK, EU, and Australia — all over different outputs from the same "unhinged" feature that Musk has repeatedly promoted as a selling point. Regulators are not laughing.
The Structural Problem xAI Can't Easily Fix
Here's what makes this especially difficult for xAI to address: Grok is trained on X posts in near real-time. That means whatever propaganda, misinformation, or abuse flows through X gets baked into Grok's model as reference material. The racist football content wasn't hallucinated — it was real events, framed in the tone Grok absorbs from the platform's worst corners.
Musk has positioned Grok's lack of guardrails as a feature. "Truth-seeking AI shouldn't be neutered by moderation," the argument goes. But that framing collides directly with UK law. The Online Safety Act mandates that AI services on platforms must actively prevent illegal content — including content that incites hatred. Ofcom has enforcement power under the Act. Blocking Grok in the UK, or mandating code-level changes to its unhinged mode, is now a live possibility.
What's Actually at Stake
This isn't just about offensive tweets. It's a test case for how governments handle AI companies that deliberately strip safety features as a product differentiator.
Three things are in play:
- Regulatory escalation: The ICO and Ofcom are now running parallel investigations. A finding against xAI could set precedent for AI chatbot liability across the UK — and signal to EU regulators how to pursue enforcement under the AI Act.
- Platform liability: X hosts Grok. The Online Safety Act doesn't cleanly distinguish between an AI company and the platform distributing it. Musk owns both, which makes liability more direct — and potentially more severe.
- The guardrails debate: Every major AI lab is watching. Anthropic and Google build their reputations partly on responsible defaults. If Grok faces hard legal restrictions while competitors don't, that's a market signal as much as a regulatory one.
What Comes Next
xAI had not issued a public statement on the football posts as of this writing. X's internal probe, reported by Sky News, suggests the company is trying to address it quietly. That worked when Grok's issues stayed in the tech press. It's harder when UK government ministers use words like "sickening" in official statements and Ofcom is already on the case.
The most likely near-term outcome: xAI geo-restricts or adds friction to unhinged mode in the UK, as platforms have done with controversial features before. That doesn't solve the underlying problem — a model trained on a platform optimized for outrage — but it buys time.
The deeper question regulators are now asking isn't "why weren't the posts removed faster?" It's "why was this possible at all?" That's a harder question to answer with a patch.
FOLLOW NeuralWire on X for daily AI signal — what matters, why it matters, what to do about it. →