-Analysis-
PARIS — Grok is the generative AI launched by Elon Musk and built into the social network X, formerly known as Twitter, which is used by hundreds of millions of people around the world. The American billionaire also recently launched Grokipedia, an online encyclopedia meant to compete with Wikipedia. It’s becoming clear how much Elon Musk is investing in areas that influence information, knowledge and learning.
It is therefore particularly worrying to see that systems like this can produce large-scale disinformation. Within just a few days, two cases have come to light showing the dangers of poorly controlled technology with safeguards applied only after the fact.
The Paris public prosecutor has expanded an investigation that already targeted the X platform to include Grok, after the AI generated answers that blatantly were Holocaust-denialism. In response to a question, Grok claimed that Nazi gas chambers were used to “disinfect Jews against typhus” rather than to kill them.The answer was later corrected following public outcry.
A few days earlier, Grok’s false statements about what happened at the Paris concert hall Bataclan on November 13, 2015 had already sparked controversy. All of this raises serious concerns.
What’s behind AI errors
There are two issues here. The first is “hallucinations,” the term computer scientists use for completely invented answers. If you have used ChatGPT or other systems, you will have noticed that responses sometimes include fabricated details.
The second issue is even more serious: the bias in the data the AI relies on. Even when hallucinations are not involved, generative AI does not create answers from scratch; it draws on the huge amount of data it has been trained on. Errors and even misinformation are part of that data.
But there are also different reference systems. In the case of the gas chambers, Grok responded to a user on X who challenged it by saying that “freedom of expression,” is protected by the U.S. Constitution, and allows for critical examination of historical narratives without prior censorship. Grok claims that labeling any questioning based on evidence as ‘denialism’ stifles debate. This is the heart of the problem.
Orwellian ways
And this is the big difference between the United States and Europe: Holocaust denial is a crime in Europe, but not in the United States, where the First Amendment has a much broader view of free speech, which these days allows “alternative truths” to spread.
Musk fully embraces this view. He recently claimed that anyone who gets their information from traditional media lives in an “alternative reality,” flipping the meaning of words in an Orwellian way. His Grokipedia encyclopedia follows the same logic.
This raises the question of regulation. In recent years, Europe has tried to equip itself with tools to regulate the digital world, especially concerning the content of large platforms.
But American pressure is now so strong that the European Commission, with support from Paris and Berlin, plans to loosen these regulations, opening the door to large-scale disinformation. Europe must not let Musk be the one who defines what the gas chambers were.