AI Chatbot Grok Accused of Spreading Misinformation About Bondi Beach Shooting

In a harrowing scene that unfolded on Australia’s Bondi Beach on the evening of December 14, what began as a Hanukkah celebration turned tragic when gunmen opened fire on the gathered crowd. The attack has claimed 15 lives as of the latest count, making it one of the most deadly incidents in Australia’s recent history.

Amid the chaos, one man emerged as an unexpected hero. Ahmed Al Ahmed, a bystander, confronted one of the attackers in a moment of extraordinary courage. Video footage captured Al Ahmed grappling with one of the gunmen, successfully disarming him by wresting away a long-barreled weapon. The assailant, dressed entirely in black, stumbled and fled after losing control of the firearm. This decisive action likely prevented an even higher casualty count.

The video of Al Ahmed’s heroic intervention spread rapidly across social media platforms, with many praising his quick thinking and bravery in the face of extreme danger. Such civilian interventions, while rare, have become more visible in an era where mass shootings are increasingly documented in real-time.

However, the tragedy took an unexpected turn in the digital realm when Elon Musk’s AI chatbot Grok, available on the social platform X (formerly Twitter), completely misinterpreted the widely circulated footage. When users asked Grok to explain the video, the AI bizarrely described it as “an old viral video of a man climbing a palm tree in a parking lot, possibly to trim it.”

This wasn’t Grok’s only misrepresentation. In separate responses to user queries about the same footage, the chatbot incorrectly labeled it as footage from the October 7 Hamas attack, and in another instance, attributed it to Tropical Cyclone Alfred, according to reporting by Gizmodo.

X has yet to provide an explanation for these significant errors, which extend beyond queries specifically about the Bondi Beach attack.

Digital misinformation experts point to a fundamental problem: AI chatbots consistently struggle with breaking news scenarios. NewsGuard researcher McKenzie Sadeghi explained the mechanism behind these failures, noting, “Instead of declining to answer, models now pull from whatever information is available online at the given moment, including low-engagement websites, social posts, and AI-generated content farms seeded by malign actors.”

This isn’t an isolated incident for Grok or other AI systems. Following the recent killing of far-right commentator Charlie Kirk, Grok amplified conspiracy theories about the shooter and Kirk’s bodyguards, with some users being told that graphic video footage of Kirk’s death was merely a meme. Google’s AI Overview similarly provided false information in the immediate aftermath of that incident.

The problem is exacerbated by the broader reduction in human fact-checking across social media platforms. AI systems often prioritize providing rapid responses over ensuring accuracy when dealing with developing news situations.

Major technology companies recognize this critical weakness in their AI offerings. In response, they’ve pursued increasingly substantial licensing deals with news organizations. Meta recently formalized multiple commercial AI agreements with prominent news publishers including CNN, Fox News, and the French publication Le Monde, adding to its existing partnership with Reuters. Similarly, Google is currently testing AI-powered features like article summaries on Google News through a pilot program with participating publishers.

Despite these efforts, AI hallucinations—instances where AI systems confidently present false information—remain a persistent challenge for large language models and chatbots. The Bondi Beach shooting response is merely the latest example of how AI systems can spread misinformation during critical events when accurate information is most essential.

As investigations into the Bondi Beach attack continue and communities mourn those lost, the incident serves as a sobering reminder of both human capacity for heroism and the technological limitations that can compound tragedy in the digital age.