Superintelligence gamble, how a ten-percent warning revives the 1939 nuclear scare

When a senior safety researcher at Anthropic, an AI safety lab, Ivan Hopingher assigns a ten-percent probability of human extinction within the next decade due to unchecked AI acceleration, the crisis exceeds the abstract figure and collides with a sharp split between declared panic and cold neglect. It is the same percentage that might lead someone to ignore carrying a umbrella in anticipation of rain, yet it turns into an existential shock demanding comprehensive review when linked to the fate of entire societies and nations.

Hopingger’s warning followed Jacob Coxon’s resignation from the Anthropic lab after a prior stint at OpenAI; in public posts he explained that large labs are rushing straight toward self-improving superintelligence and risking people’s lives. This coincidence prompted international analysts, including Yusi Miklberg, a fellow at Chatham House’s Middle East and North Africa program, to invoke a pivotal historical moment from August 1939, when Leo Szilard and Albert Einstein sent their famous letter to U.S. President Franklin Roosevelt warning that uranium research was not limited to energy generation but opened the way to weapons of massive destructive power.

The historical episode shows that technology is inherently ethically and socially neutral, and that its creators and regulators determine whether it becomes a force for benefit or a tool of harm.The most urgent danger appears in stripping daily decisions of human judgement.Before any assumed catastrophic scenario unfolds, unchecked rush toward automating office tasks such as document review and letter drafting deprives thousands of young people of their first jobs, generating sharp unemployment and a societal feeling of bitterness and anger toward technological change, an effect that far exceeds what earlier waves of digitisation have caused.

This regulatory concern expands to sensitive sectors including health care, where algorithms promise disease-diagnosis accuracy while risking the erosion of humane clinical judgment, as well as public services such as air traffic control and the concentration of power in the hands of a few that operate the most robust systems. Yet the deeper shift emerges on military battlefields, where autonomous weapons pose a complex ethical and legal dilemma in the event of a lethal error targeting objectives without direct human supervision: who bears criminal responsibility then? The programmer, the field commander, the manufacturer, or the state? The geographic distance from the battlefield lowers psychological and political barriers to weapon use and paves the way for rapid clashes between machines interacting with machines beyond any human restraint.

This debate is no longer a theoretical luxury for the West; it directly touches operating costs and supply strategies in the Gulf, Egypt and the Levant. When out-of-control warnings come from safety engineers within labs that sell models to regional banks and their public entities, reliance on closed-source models becomes a regulatory risk requiring independent audit.Regional institutions must build local inspection and governance layers that keep humans supervising decisions., while protecting the career paths of young staff from rapid erosion due to the elimination of entry-level office jobs, and imposing strict oversight that prevents delegating automated systems in critical and security infrastructure without a decisive human review.

The editor-in-chief confirms that the solution does not lie in halting scientific research nor in reckless rushing without brakes, but in pacing publication so that societies and legislators have genuine time to examine the limits of AI use and its regulation. The real stake is not about which side wins the human-machine race, but about societies’ ability to harness emerging tools without losing control over the social system the technology was created to serve.