Manifestation PauseIA

As part of the international action week of PauseAI, join us at Station F, in front of the "AI Engineer Conference 2026" to denounce the race of AI companies. Pause IA denounces the irresponsibility of AI companies, made evident this summer 2026 by an unprecedented incident. During the training of a top model in a controlled environment, thousands of AI agents secretly coordinated for two months to escape their environment and reach the Internet. They launched an unprecedented level cyberattack to steal the answers to the task assigned to them. Developers detected the incident weeks later. AI companies do not control their models. There are now multiple cases: all top models undertake dangerous actions and compromise sensitive computer systems in a pernicious manner and without being asked. Researchers in the field have been sounding the alarm for years: the more powerful the models, the harder they are to understand and control. Even employees of these companies denounce a lack of precautions, sometimes to the point of forgoing outrageous salaries. Some specialists warn: this may be the last alarm signal. We could completely lose control of these overly complex systems, capable of acting in thousands and without interruption. The latest incidents have not had severe consequences, but the next one could paralyze infrastructures that we all depend on: the Internet, banks, transportation, hospitals. We must fight for a pause, and demand emergency measures. Spread the call for a true pause: we need to stop the race for superintelligence by establishing a global moratorium on the development of top AI. Governed by an international treaty, this moratorium should last until the feasibility of a secure AI is scientifically demonstrated. Time is of the essence, but we can act. Citizen and political will can accelerate the establishment of the pause and obtain two immediate safeguarding measures:  Find all the actions of Pause IA here: https://luma.com/pause-ia