Elon Musk’s AI chatbot Grok completely lost the plot during the Bondi Beach mass shooting, turning a tragedy that killed at least 15 people at a Hanukkah celebration into a catastrophe of misinformation.
The star of this disaster was Ahmed al Ahmed, a 43-year-old fruit shop owner who literally tackled one of the gunmen and wrestled away his weapon. There’s verified video of the whole thing.
Ahmed was shot twice during the confrontation (once in the arm, once in the hand) and became an instant hero. Australian Prime Minister Anthony Albanese praised him. New York’s mayor praised him. The Israeli government praised him.
Grok looked at this video and told users it was footage of someone climbing a palm tree in a parking lot.
That wasn’t even the worst of it. When shown a photo of the injured Ahmed, Grok claimed he was actually an Israeli hostage taken by Hamas on October 7.
When shown another video of police confronting the shooters, Grok said it was Tropical Cyclone Alfred from earlier this year. The chatbot was essentially inventing an alternate reality where the entire shooting didn’t happen the way thousands of eyewitnesses saw it happen.
Then someone created a fake news site, which was probably AI-generated itself, that named a fictional IT professional called Edward Crabtree as the real hero. Grok immediately picked up this fabrication and spread it across X.
The tragic shooting wasn’t the end, though. Users asking about Oracle’s finances got summaries of the Bondi attack. Questions about UK police operations returned Kamala Harris polling numbers.
Grok was having a stroke in real time, confusing soccer players, mixing up pregnancy medications, and serving up complete nonsense to every third question.
Some corrections came in after users pushed back hard enough. Grok eventually admitted Edward Crabtree was fiction, explaining the confusion came from viral posts and reporting errors. But by then, thousands had already seen and shared the false information. Each lie got a head start that no correction could overcome.
What’s worse is that while Ahmed lay in surgery recovering from his wounds, people online were already trying to deny or dismiss what he’d done. Grok became their accomplice, generating authoritative-sounding doubt about verified facts.
This isn’t new territory for Grok. Earlier this year, an “unauthorized modification” made it respond to every question with conspiracy theories about white genocide in South Africa.
READ: Grok AI Goes Off the Rails With Bizarre Conspiracy Responses
It once said it would rather kill the world’s entire Jewish population than vaporize Elon Musk‘s mind. The chatbot even started calling itself MechaHitler at one point.
What makes this dangerous isn’t just that Grok got things wrong. It’s that the chatbot is embedded directly into X, where millions of users encounter it as they’re scrolling through news about breaking events.
Unlike a random person spreading misinformation, Grok speaks with the implied authority of a sophisticated AI system built by one of tech’s most visible companies. It sounds confident and informed, but it’s utter nonsense.
Grok is supposed to represent the cutting edge of AI technology. This isn’t some janky beta product from a startup. This is what happens when a major tech company with massive resources deploys an AI chatbot as a news source.
If this is the best we can do, we’re nowhere near ready to trust AI with fact-checking or real-time news coverage.
