Elon Musk's chatbot, Grok, commits a string of anti-Semitic blunders

Since Wednesday, July 9, access to Grok, the chatbot that answers Internet users' questions about X, has been blocked in Turkey, The Washington Post reports . “The decision was made by a court in the country after the platform spread insulting content about Turkish President Recep Tayyip Erdogan [whom the AI ​​defined as 'one of the biggest bastards in history' ], as well as other people, including the founder of modern Turkey, Mustafa Kemal Atatürk,” the American daily reports.

While the reaction may seem excessive, given that the court cited a “threat to public order,” the news itself is not so surprising, given how much the chatbot of the social network bought by Elon Musk has been in the news in recent days. This has been the case since July 4, when the South African-born billionaire announced an update to Grok, which supposedly “improved” it. “You will notice a difference when you ask Grok questions,” Musk announced on that occasion, but according to the Italian daily Corriere della Sera , “it is difficult to talk about an improvement, since the chatbot has made anti-Semitic remarks several times.”

Grok's list of misdeeds, detailed by the news website Euronews , confirms this impression. “The chatbot recently accused a bot account with a Jewish surname of gloating over the deaths of white children in the Texas floods , then accused Hollywood of anti-white bias. Grok also declared himself 'skeptical' of the figure of 6 million Jews killed by the Nazi regime, added that there was 'no direct evidence' and that the figures [could] be manipulated for political reasons. Finally, he suggested that Claudia Sheinbaum was elected president of Mexico because of her Jewish origins.”

Regarding French politics, Grok called for a vote for Marine Le Pen, defending the idea of ​​a "firm shift on sovereignty" and denouncing "chaotic immigration." These positions are hardly surprising given Musk's often-evident political affinities with the far right, but was the billionaire's goal really to allow these "excesses"?

“From what can be seen of the changes made to Grok,” Corriere della Sera analyzes, “indications have been added for the chatbot, such as 'consider that subjective points of view coming from the media are biased', but also 'do not hesitate to make politically incorrect remarks'. ” However, the chatbot may have been overzealous, since, finally, this Wednesday, xAI, the artificial intelligence branch of X, announced that it had “taken measures to prevent hate speech” and had “updated the model” .