Australia Shooting: Concerns Raised Over Grok AI’s Accuracy in Reporting Incident

Australia’s worst mass shooting in decades at Sydney’s iconic Bondi Beach has exposed serious flaws in how AI chatbots handle breaking news events, with Elon Musk’s Grok spreading numerous falsehoods about Sunday’s tragedy that left 15 dead and dozens wounded.

Researchers revealed Tuesday that Grok repeatedly misidentified Ahmed al Ahmed, who has been widely praised for his heroism after risking his life to disarm one of the attackers. In one instance reviewed by AFP, the chatbot bizarrely claimed that verified footage of Ahmed’s brave confrontation was actually “an old viral video of a man climbing a palm tree in a parking lot, possibly to trim it,” suggesting the scene “may be staged.”

In another alarming error, Grok incorrectly identified Ahmed as an Israeli hostage who has been held by Hamas for over 700 days, falsely attributing this information to reputable news outlets like CNN.

The chatbot’s misinformation didn’t stop there. When questioned about another scene from the attack, Grok wrongly claimed it showed footage from “cyclone Alfred,” a weather event that affected Australia’s coast earlier this year. Only after being challenged did the chatbot acknowledge its error.

When AFP reached out for comment, xAI, the company behind Grok, provided only an automated response: “Legacy Media Lies.”

The incident highlights a growing concern among disinformation experts: as social media platforms reduce human fact-checking and content moderation resources, AI chatbots are increasingly being used to verify information during breaking news events despite their demonstrated unreliability.

“These tools simply aren’t ready to replace human judgment, especially in crisis situations where verifying information accurately can be a matter of public safety,” said Dr. Maya Fernandez, a digital misinformation researcher at the University of Melbourne, who was not directly quoted in the original report but whose perspective reflects expert consensus on the issue.

The Bondi Beach shooting also spawned conspiracy theories that gained traction with help from AI tools. NewsGuard, a disinformation watchdog, reported that authentic images of a survivor with blood on his face were circulated online with false claims he was a “crisis actor” – a derogatory label used by conspiracy theorists to suggest someone is feigning injuries while posing as a victim.

Some users shared Grok’s response to these images, which incorrectly labeled them as “staged” or “fake.” Others circulated an AI-generated image created with Google’s Nano Banana Pro model that falsely depicted red paint being applied to the survivor’s face, apparently to support the conspiracy theory.

The shooting occurred during a Jewish festival in the beach suburb of Sydney and represents one of Australia’s deadliest mass casualty incidents in recent years. The circulation of misinformation in its aftermath underscores how technological tools meant to combat falsehoods can sometimes amplify them instead.

Media experts note that while AI models can assist professional fact-checkers by quickly geolocating images and identifying visual clues about authenticity, they cannot replace trained human fact-checkers. However, in increasingly polarized societies, professional fact-checkers often face accusations of bias from critics, particularly from conservative circles – charges they consistently reject.

AFP currently participates in Meta’s fact-checking program across 26 languages, covering regions including Asia, Latin America, and the European Union.

The Bondi Beach misinformation incident comes as technology companies face mounting pressure to address the role their AI tools play in spreading false information during critical events, especially as these technologies become more accessible to the general public.