AI Insiders Warn of 10% Chance of Human Extinction Within a Decade

High-level warnings from artificial intelligence industry insiders have reignited urgent debates over the existential risks posed by frontier models. The conversation has shifted from academic speculation to a pressing policy crisis as those building the technology warn that humanity could face extinction by the end of the decade.

Evan Hubinger, the head of alignment science at Anthropic, has estimated there is a greater than 10% chance that AI could kill all humans within ten years. This sentiment is not isolated to a few alarmists; a 2024 survey of nearly 3,000 AI researchers revealed that more than half believe the probability of AI causing human extinction or severe disempowerment—a metric known as 'p(doom)'—is at least 10%. Jacob Coxon, a former researcher at Anthropic, further emphasized the gravity of the situation upon his resignation, stating that the people actually building AI "earnestly believe that it could kill us all by the end of the decade."

The Alignment Crisis

The current surge in alarm is driven by the rapid emergence of 'frontier' models and an intensifying competitive race between Chinese labs and U.S. firms like OpenAI and Anthropic. At the heart of the fear is the problem of 'alignment': the technical challenge of ensuring that a superintelligent system's goals remain compatible with human survival. As systems grow more powerful, the risk increases that they may pursue objectives that inadvertently lead to human catastrophe.

This tension is compounded by internal cultural divides within the industry. Jaan Tallinn has estimated that 10-15% of AI employees believe the technology will eventually serve as a worthy successor to humanity, suggesting that some developers view the displacement of the human species as an acceptable or inevitable outcome.

Geopolitical and Legislative Stakes

If the 10% risk estimate is accurate, AI represents a systemic threat on the same scale as nuclear war or catastrophic climate change. The current trajectory suggests a dangerous friction between commercial profit, geopolitical dominance, and global safety. Without international coordination or a deliberate 'pacing' of development, critics argue that humanity may lose control of its most powerful technology before any viable safety frameworks can be established.

These risks have already triggered legislative action. U.S. Senator Bernie Sanders has pushed for strict regulation, arguing that a handful of "greedy people" cannot be allowed to "play God and determine the future of humanity." To this end, Sanders has introduced the 'Ban Artificial Superintelligence Act,' a legislative effort designed to prevent AI from developing beyond human control.

The Path Forward

While some industry leaders, such as Hugging Face CEO Clement Delangue, have dismissed these warnings—comparing the act of asking a researcher about extinction to asking an air conditioning technician about climate change—the momentum is moving toward formal treaties. The coming years will determine whether the international community can establish a binding framework to limit the creation of artificial superintelligence or if the competitive race will override the precautions demanded by the scientists building the systems.