00:00How X's AI chatbots struggle to fact-check the Israel-Iran conflict.
00:08Users have been turning to AI chatbots to inform themselves about the 12-day conflict between Israel and Iran,
00:16which now seems to have subsided.
00:18But a new study shows that Grok, the free-to-use chatbot developed by Elon Musk's XAI,
00:24has been misleading users with inconsistent and incorrect claims.
00:30This video, claiming to show the destruction from a retaliatory Iranian attack on Tel Aviv airport, is AI-generated.
00:38We here at Iroverify have previously debunked it.
00:42But when asked to verify its authenticity, Grok replied that it was likely real footage, not AI-generated.
00:48At Iroverify, we identified several more instances in which Grok falsely describes AI-generated videos as authentic.
00:58We also detected instances of Grok linking completely unrelated videos to the conflict.
01:04Grok describes this video, which shows festival-goers at Hellfest in France,
01:10as showing Israelis fleeing the conflict at the Taba border crossing with Egypt.
01:15And Grok also alleges that this video of an explosion in Malaysia shows an Iranian missile hitting Tel Aviv on June 19th.
01:24Researchers at the Digital Forensic Research Lab say that Grok grounds its responses on media reports
01:30that quickly become outdated due to the fast-paced nature of the conflict.
01:35The chatbot also consistently fails to corroborate the authenticity of sources.
01:39For example, it's claimed that accounts that are spreading disinformation on X are official, governmental or military accounts.
01:48Experts say users are increasingly reliant on these tools for fact-checking.
01:53Yet, these tools are in fact amplifying disinformation and conspiracies.
Comments