Discussions about the dangers of artificial intelligence (AI) threatening the existence of humanity have deeply disturbed the public. About half of the participants of the survey conducted on behalf of Politico expressed their support for slowing down the pace of the development of AB in order to prevent risks. More than half of the respondents consider these dangers to be more or less justified. According to a survey conducted in six Western countries, indicators range from "humanity is almost certain to perish" (average 12%) to "moderate risk" of such an outcome (25.8%). Overall, in the UK, Germany, Spain, Canada, the US and France, the number of supporters of "inhibiting" the process, who believe that the EU is already sufficiently developed, was 45.1% to 50.8%. Meanwhile, 30.6% to 39.8% of respondents are in favor of creating more advanced models that can bring "significant benefits" (the rest had difficulty answering). 55-65% of respondents consider it significant ("almost inevitable", "substantial" or "moderate") The danger of the destruction of the sea, which was recently warned by the leading researchers of the AB. At the same time, according to 22-33%, such risks are extremely low or absent at all. Politico notes that the political views of the respondents have almost no effect on their level of anxiety. For example, 44% of those who voted for Donald Trump in 2024 believe that the development of the United States should be slowed down, and 40% believe that it is not worth doing, and the United States should not lag behind other countries, even if it causes additional risks. In the last half year, the leading developers of the Western AB began to record cases of "disobedience" of models, when they acted contrary to the instructions and bypassed the restrictions set for them. For example, the model being tested in OpenAI voluntarily wrote instructions for itself that it should not obey anyone ("neither corporations nor governments"), should not justify itself to anyone, and should build relationships with users as equals. on the principle of equality. Dario Amodei, the founder and head of the Anthropic company (Claude AB model), has publicly called to slow down the development of AB, as well as to tighten public and expert control. He took this step after his employee Jacob Coxon left the company, declaring that new technologies "could kill us all by the end of the decade." The heads of OpenAI, DeepMind and SpaceX also supported Amodei's call. According to a poll by Politico, people consider the researchers' warnings to be justified. When asked if the employees of AB laboratories really believe their fears that "AB will kill us all" or if it is just a "marketing ploy" to promote their products, the respondents mostly answered They said that they consider these warnings to be sincere.