WASHINGTON: Elon Musk’s AI chatbot Grok circulated significant misinformation about Australia’s recent Bondi Beach mass shooting.
Researchers reported on Tuesday that Grok repeatedly misidentified Ahmed al Ahmed, a man hailed as a hero for disarming an attacker.
The chatbot falsely claimed verified footage of Ahmed’s confrontation was an old viral video of a man climbing a palm tree.
Grok separately misidentified an image of Ahmed as that of an Israeli hostage held by Hamas for over 700 days.
It also incorrectly labelled another scene from the attack as footage from tropical “cyclone Alfred”.
The AI only acknowledged the footage was from the Bondi Beach shooting after a user pressed it to re-evaluate.
xAI, Grok’s developer, responded to an AFP request for comment with an auto-generated reply stating “Legacy Media Lies.”
Online users also falsely labelled an authentic image of a survivor as a “crisis actor”, a derogatory term used by conspiracy theorists.
Grok contributed to this by falsely labelling the survivor’s photo as “staged” or “fake”.
NewsGuard reported some users circulated an AI-generated image created with Google’s Nano Banana Pro model to bolster the false claim.
The image depicted red paint being applied to the survivor’s face to pass as blood.
Researchers say this episode highlights the unreliability of AI...
As false and exaggerated information claiming that school zone speed limits will be strengthened starting next year spread online, the National Police Agency directly issued a ‘fact-check’ press r...