The artificial intelligence program ‘Grok,’ developed by Elon Musk’s xAI, disseminated false information regarding the mass shooting incident at Bondi Beach in Sydney, Australia. The attack targeted a gathering celebrating the Jewish holiday of Hanukkah on Sunday evening.
The incident, which Australian authorities have described as both ‘anti-Semitic’ and ‘terrorist,’ sparked a surge of information and images online, particularly on the X platform, where ‘Grok’ is widely used.
Initially, ‘Grok’ misidentified Ahmed Al-Ahmad, who has been hailed as a ‘hero’ in Australia after video footage emerged showing him disarming one of the attackers. Al-Ahmad remains hospitalized with serious injuries. ‘Grok’ falsely claimed the video depicted ‘an old video widely circulated showing a man climbing a palm tree in a parking lot,’ referring to a ‘staged event.’ The AI further incorrectly identified Al-Ahmad, who is of Syrian origin, as an Israeli prisoner held by Hamas for over 700 days.
When questioned about another scene from the attack, the program incorrectly asserted that it depicted footage from ‘Cyclone Alfred,’ which caused severe weather along the Australian coast earlier this year. The AI only corrected itself and acknowledged that the footage was from the Bondi Beach shooting after a user pressured it to reassess its response.
On Sunday evening, a man and his son opened fire on a crowd gathered at Bondi Beach to celebrate Hanukkah, resulting in at least 15 fatalities and 42 injuries.
In an automatic reply to an AFP inquiry, xAI, the program’s developer, stated, ‘Traditional media lies.’
Experts believe that AI can be beneficial for tasks such as geolocation of images, but they maintain that it is currently impossible to replace human verification processes entirely. The incident highlights the dangers of relying solely on AI for information verification, especially in sensitive situations involving breaking news. Further development and refinement of AI fact-checking capabilities are crucial to prevent the spread of misinformation.
