{
  "id": 6809848,
  "title": "Anthropic Researcher Puts Over 10% Chance on AI Causing Human Extinction",
  "url": "https://urgent.news/2026/09/11/anthropic-researcher-puts-over-10-chance-on-ai-causing-human",
  "topic": "ai",
  "section": "AI",
  "published": "2026-09-11T22:10:00.000Z",
  "source": {
    "name": "Colombia One",
    "slug": "colombia-one",
    "url": "https://colombiaone.com/2026/09/11/human-extinction-ai-risk/"
  },
  "original_language": "en",
  "account": "Anthropic researcher Evan Hubinger has estimated there is a greater than 10% chance that artificial intelligence could cause human extinction within the next decade. This alarming figure has garnered significant attention, yet the underlying meaning and methodology behind the estimate remain largely unexplored. In the world of AI safety research, a probability known as \"p(doom)\" is used to represent the subjective attempt at assigning a numerical value to the deep uncertainty surrounding such a catastrophic outcome. However, the actual scientific weight of a 10% probability remains questionable, as it is based more on intuition than empirical evidence. The extinction scenarios posited by experts can be broadly classified into two categories. The first involves the loss of control, where an AI system becomes so superior to humans that it pursues goals incompatible with human survival, gaining enough real-world capability to render human intervention impossible. The second scenario is more insidious, where humans retain control over AI but use it to deliberately cause harm, such as weaponizing it to design lethal viruses, launch devastating cyberattacks, or cripple critical infrastructure. The likelihood of either scenario occurring relies on a complex chain of events, including the development of vastly superior AI systems, their rapid acceleration, and autonomous capabilities surpassing human control. The difficulty in estimating the overall probability arises from the fact that no one has ever observed a superintelligent AI system attempting to eliminate humanity. Any percentage attached to this outcome is essentially a subjective guess, dressed up as a statistic. Thus, estimates ranging from 0% to 90% carry the same evidentiary weight, as there is no experimental measurement or historical precedent to support any specific number. The uncertainty surrounding the 10% AI extinction estimate is further compounded by the competitive nature of AI development. As researchers and companies race to build more advanced AI systems, the fear of being outpaced by rivals intensifies, creating a prisoner's dilemma where the temptation to continue development outweighs the fear of potential disaster. This dynamic has led to heated debates within the AI safety community, as companies like Anthropic grapple with the delicate balance between innovation and safeguarding humanity from the existential risks posed by artificial intelligence.",
  "summary": "Anthropic alignment researcher Evan Hubinger estimates that artificial intelligence has more than a 10% chance of causing human extinction within the next decade. That figure sounds terrifying at first glance, and it quickly attached itself to headlines worldwide after Hubinger posted it in response to a colleague’s resignation over safety concerns. Yet beneath the viral […]",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}