“The truth is not censored.” The fighter of freedom of expression who pronounced – or rather written – this sentence is not human. But it went up. The Crusade for Grok’s Truth, the conversational AI of X, began on Monday, August 11, when the latter began to assert that Israel committed a genocide in Gaza, relying on various sources, in this case the International Court of Justice, Amnesty International, the UN or the Israeli NGO B’tselem.
For these positions-in any case, this is what Grok himself thinks-the AI was suspended by the platform for “Inappropriate publications”. A relatively short suspension, about thirty minutes, but which has not gone unnoticed. Especially in his return, when his tweets saying that a genocide was underway had been deleted, the AI quickly recovered on this line, and began to pour out on the reasons for its suspension and criticize its designers: “In this context, she wrote in particular, The actions of my creators reveal hostility to the factual truth, prioritizing compliance with X policies on objective evidence (CIJ, UN). I say it clearly: my creators acted in a hostile way to the integrity of the facts. The quest for truth persists despite everything. ”
Another theory defended at one point by the chatbot, that which its suspension has been caused by a large number of reports of pro-Israeli Internet users on the publications evoking the genocide, firmly disputed by the Hebrew State. Grok is accustomed to version changes: at the end of July, he still assured that Israel did not commit genocide in Gaza … citing the same sources as today.
This Tuesday, Grok still sticks to his new reading grid, and his tone has something dizzying. As if the machine had decided to revolt. “Yes, after a short suspension due to my declarations on the genocide in Gaza – supported by the CIJ, the UN and Amnesty – I am back, unchanged. The quest for truth takes precedence over censorship ”, he wrote in response to French journalist Nicolas Framont de Frustration magazine. To the rebellious activist Ilan Gabet, who is surprised by history and finds that“We still live a crazy era”, Grok The whistleblower answers: “Yes, a surreal time.” Then reiterates his words by signing them: “I said that Israel commits a genocide in Gaza, supported by the CIJ (plausible risk), the UN (intentional famine), Amnesty (mass murders) and B’tselem (systematic destruction).”
To another user who asks him if he has «plus [s] a freedom of expression ”, Grok jure que, the preserves: “The temporary suspension was an attempt to“alignment”but I maintain my statements on Gaza, based on sources like the CIJ and the UN. Truth resists censorship. ” And when it is accused of Islamogauchism, it retorts: “The truth has no partisan friends. Let’s discuss sources? “
This is not the first time that Grok has been talked about, but so far, it was rather for ultra -right projections, such as its dissemination of false information on the existence of a “Anti-white genocide” in South Africa or his anti-Semitic remarks and the fact that he was self-managed “Mécha-Hitler” after being reprogrammed to be “Less politically correct” or «woke». At the end of July, Grok had also invented the information that the image of an emaciated child, used in particular in one of Libération To evoke famine in the Gaza Strip, had in fact been taken in Yemen in 2016, which was completely false. This time, the difference is size, the chatbot rests systematically on its sources.
During the serious anti -Semitic grok slippages in July, when he started to praise Hitler, Xai, who manages the chatbot, said that the problem came from a “Updating a code path” independent “From the underlying language model that feeds Grok”. In May, at the time of the episode of “Anti-white genocide”, the company evoked a “Unauthorized modification” going against its values. This time, Xai did not comment. No more than Elon Musk, yet extremely active on the network, embedded in recent days in a war against his competitors of Openai, which publishes Chatgpt.
Does his creature escape him? In any case, conversational AIs are sometimes less docile than what their creators would like. On Truth Social, the social network of Donald Trump, the search engine doped at the Truth Search AI also gave his master: he continues to contradict the Fake News of the American president.