AI: A study highlights Grok’s deficiencies in his answers to the Iran-Israel War

Grok, the artificial intelligence assistant (AI) of the new company of Elon Musk XAI, has generated inaccurate and contradictory responses to requests to verify information about the 12 -day war between Israel and Iran, according to a study.

With the gradual deletion of the verification made by journalists on platforms such as Facebook, more and more Internet users resort to artificial intelligence tools such as Grok to find reliable information, but the responses obtained sometimes expose them to misinformation.

Grok loses reliability

According to the study published on Tuesday of the Digital Laboratory of Forensic Research (DFRLAB) of the Atlantic Council, a circle of American reflection, “Grok’s actions during the first days of the conflict between Israel and Iran revealed significant defects and showed the limits of the chatbot to provide precise, reliable and coherent information in crisis times.”

He analyzed 130,000 messages in several languages ​​on the X platform, which integrates the Chatbot Grok, and determined that the latter “had trouble authenticating the content generated by AI.”

For example, after Iranian retaliation attacks in Israel, Grok proposed very different responses to the instructions similar to the images generated by the AI ​​of an destroyed airport, which had raised millions of views on X: they ranged, sometimes in the same minute, between denying the destruction of the airport and confirmed that it had been beaten by attacks.

And in some answers, Grok advanced that a Yemen missile was at the origin of the damage, while in others, he falsely identified that the airport in question was in Beirut, Gaza or Tehran.

According to specialists, Grok has already been caught in his facts to verify events during the recent conflict between India and Pakistan and during the manifestations in Los Angeles against Donald Trump’s anti-immigration policies.

Last month, the AI ​​service of the South African businessman Elon Musk had also been controversial by erroneously evoking a “white genocide” in South Africa, an error that Xai charged by XAI to an “unauthorized modification.”