In an unexpected turn within the tech sector, Anthropic has decided to pause the public development of its advanced AI system, Claude Mythos Preview, after considering that its capabilities pose a significant risk. According to the company itself, the results obtained by Mythos were deemed "too dangerous for the general public." The initial goal of Mythos was to identify vulnerabilities in systems like Linux, but its effectiveness exceeded expectations to the point that it could facilitate cyberattacks rather than protect against them. Anthropic's decision has been to restrict the use of Mythos to a defensive cybersecurity program, limited to a select group of partners. The main reason: the model proved capable of surpassing the company's own security barriers, detecting critical flaws in the Linux kernel, the core of most global servers. This opens the door for malicious actors to take control of key infrastructures. Additionally, Mythos identified a 27-year vulnerability in OpenBSD, an open-source operating system used in high-security environments and critical infrastructures. This weakness would allow attackers to block any machine running it. Anthropic values Mythos at over $900 billion, placing it above OpenAI as the most valuable AI startup. Jack Clark, co-founder of Anthropic, estimates there is more than a 60% chance that, before 2028, an AI capable of training its own successor without significant human intervention will emerge, a scenario he finds concerning and for which society is unprepared. Max Tegmark, a physicist at MIT and founder of the Future of Life Institute, has been warning for years about the risks of an uncontrolled AI race. In 2023, he promoted an open letter signed by over 33,000 people, including Elon Musk, calling for a pause in the development of advanced AI. Tegmark warned about the lack of regulation and the dangers of leaving technological advancement in the hands of corporate interests, recalling precedents like thalidomide or asbestos. The Mythos crisis has had immediate repercussions. The White House, traditionally reluctant to intervene in AI development, is now considering an executive order to oversee and, if necessary, veto new models before their release. Under the Trump administration, the strategy had been "zero regulations" to compete with China, but the National Security Agency (NSA) has already accessed Mythos for testing. In this context, two AI experts, Dean Ball and Ben Buchanan, published a joint article in the New York Times emphasizing the need for political unity in the face of AI risks. They point out that China could use tools like Mythos to exploit vulnerabilities in U.S. systems and that, although the West maintains a technological advantage, it is fragile. The United States and its allies have a processing capacity far superior to that of China, but international cooperation will be essential to manage catastrophic risks, such as the use of AI in biological weapons. In Spain, concern is growing over the vulnerability of critical infrastructures. The Spanish government has requested early access to Mythos to assess risks in the European financial sector, although Spain's position in NATO and its relationship with China complicate its access to sensitive technologies. Amid this uncertainty, Anthropic has announced a partnership with Blackstone, Hellman & Friedman, and Goldman Sachs, with an investment of $1.5 billion, to integrate Claude into critical operations of medium-sized companies and, later, small ones globally. The debate over the future of AI and its impact on society remains open, with scenarios reminiscent of science fiction dystopias.