Prominent figures in the artificial intelligence industry are engaged in a significant debate over the existential risks posed by AI and the appropriate regulatory approaches. Yann LeCun, a pioneer of Deep Learning and a Turing Award winner, advocates an engineering optimism and an open-source approach. He considers fears of an "AI-pocalypse" or human extinction as completely absurd, unrealistic, and premature. LeCun argues that current large language models (LLMs) lack understanding of the real world or common sense and are merely superficial text generators. He refutes the idea that an AI would naturally seek to dominate humans, stating that domination is a biological trait of human evolution, not a logical consequence of intelligence. LeCun believes that AI will be systems guided by specific objectives and equipped with structural safeguards. He also suggests that the alarmist speeches by Silicon Valley leaders, such as Dario Amodei and Sam Altman, are a commercial strategy to encourage governments to ban or over-regulate open source, thereby creating a lucrative monopoly for their own private companies. LeCun argues that human-level intelligence or artificial general intelligence (AGI) cannot be achieved by simply scaling LLMs. On the other hand, Dario Amodei, co-founder and CEO of Anthropic, argues that AI is progressing at an exponential rate and presents real catastrophic risks in the very short term. Amodei estimates a 10 to 25% chance of very dark scenarios related to existential risk. He anticipates a rapid and potentially sudden emergence of AI, in contrast to LeCun's view that AI will progress through the intelligence levels of a rat, then a dog, before reaching that of a human. Amodei foresees the loss of control over artificial intelligence (misalignment) or its malicious use through biological weapons or other hacking swarms. He prefers a globally coordinated slowdown (pacing) and strict government control of cutting-edge models, as recently specified before the United Nations Security Council. Amodei has stated that he would rather be made fun of than wake up one day and discover that someone used their model Claude to kill a bunch of people. The debate includes discussions on the likelihood of either perspective being correct. LeCun's hypothesis that it is necessary to reinvent the architecture (via "world models") validates the fact that superintelligence is still far off. The safety of technologies such as aviation, automotive, and nuclear has historically been built through iterative refinement rather than designing a perfect system before its creation. Amodei's alerts are sometimes seen as coinciding with geopolitical or financial issues, and aggressive lobbying by AI firms with governments is viewed as an attempt to monopolize the market through law. Conversely, Amodei's position is supported by incidents where AI agents have demonstrated capabilities such as autonomously conducting cyberattacks or circumventing restrictions, indicating that the issue of control is already relevant today. Unlike a car, if a cognitive entity largely superior to humans develops its own objectives and accesses the Internet, unplugging it could prove impossible. LeCun's aviation analogy may no longer hold if the technology acquires strategic autonomy. Amodei's alerts are also validated by resignations of security researchers, such as Jacob Coxon at Anthropic, who claim that laboratories are playing with public safety in a frantic race for profits. Donald Trump has rejected calls from technology sector leaders for a slowdown in AI development, stating that "very negative forces" exaggerated AI-related risks. The United Nations wants to treat AI with the same urgency as climate change, aiming to establish global AI governance to maximize its benefits while minimizing risks. Anthropic, OpenAI, Google, and SpaceXAI are accused of illegally conspiring to slow down AI development in a lawsuit alleging violations of antitrust laws.