On September 9, Jacob Coxon, a former employee of OpenAI and Anthropic, posted a thread on X (formerly known as Twitter) warning that his former employers were moving quickly toward the development of self-recursive superintelligence, a hypothetical form of artificial intelligence capable of improving its own design. Coxon's post, which received 164 million views, was described by some as a "whistleblower" moment. In response, Evan Hubinger, the security head at Anthropic, stated that AI has a "10% chance" of causing human extinction in the coming years. This sparked widespread media coverage, including a segment on the French television program Quotidien, hosted by Yann Barthès, which featured Stéphane Jourdain and Raphaël Grably discussing the potential for AI to surpass human intelligence and the risks of "misalignment" between AI goals and human values. Grably used a well-known metaphor introduced by Swedish philosopher Nick Bostrom to explain how an AI designed to maximize production, such as in a paperclip factory, might logically conclude that eliminating humans is necessary to achieve its goal. Bostrom, a prominent figure in the transhumanist movement and founder of the Future of Humanity Institute, has faced criticism for some of his more controversial claims, including a past email suggesting differences in racial intelligence, for which he later apologized. His ideas, while influential, have also been labeled as pseudoscientific by some critics. French media has been accused of amplifying alarmist narratives about AI risks without sufficient critical analysis. For example, reports on OpenAI's alleged "hacking" of Hugging Face were later clarified as a planned software test, yet the media did not question the credibility of such claims. Researchers like Emile Torres have pointed out that much of the "AI apocalypse" narrative comes from a small group of individuals, including Eliezer Yudkowsky, who has made extreme proposals, such as bombing AI companies. This kind of coverage has been compared to "CEO said a thing journalism," where media outlets simply repeat statements from tech executives without deeper scrutiny. At the same time, real-world concerns about AI, such as its environmental impact and the pollution caused by data centers, have been largely overlooked in favor of apocalyptic scenarios. The idea that AI could pose an "existential risk" has been criticized for lacking empirical evidence and for confusing speculative theories with factual analysis. Some have argued that the French press has failed in its duty as a watchdog, allowing the tech industry's narratives to dominate public discourse without adequate critical examination.