A leading AI developer warns that artificial intelligence could pose a real danger to humanity. Recent statements have triggered urgent calls to slow down AI progress. The debate now centers on how real these dangers are and what action is needed.
Evan Hubinger, a researcher at Anthropic, has put a number on a fear that has been simmering in the tech world: he estimates there is more than a 10% chance that artificial intelligence could wipe out humanity within the next ten years. This prediction has pushed the issue of AI safety into sharper focus and led to renewed calls for slowing down development.
Hubinger’s warning comes amid a series of similar statements from AI researchers, who are increasingly uneasy about how quickly these systems are advancing. The concern is not just about smarter algorithms, but about the pace at which AI is gaining new abilities—possibly faster than oversight and safety measures can keep up.
Anthropic CEO Dario Amodei has publicly called for slowing the pace of frontier AI development, arguing that safety measures are not keeping up with the rapid growth in model capabilities.
As developers and policymakers weigh the risks, calls to slow AI research have grown louder. Hubinger’s scenario—where AI systems become so powerful and independent that they could threaten humanity—has moved from science fiction to a real topic of debate among those building the technology.
Specific numbers are rare in discussions about AI risk, but Hubinger’s estimate has added urgency. His claim of a greater than one-in-ten chance of human extinction from AI within a decade is not easily dismissed. It has already shaped policy conversations and research priorities around the world. According to The Guardian and other major outlets, at least two other current or former Anthropic employees have echoed this estimate, suggesting a growing concern within parts of the industry that the risk is both real and near-term.
Some experts say these predictions are speculative, but the fact remains that leading figures are now openly discussing worst-case scenarios. This shift has led to more proposals for stricter oversight, international cooperation, and even temporary pauses on the most advanced AI projects. The debate has shifted from whether AI could be dangerous to how soon and how severe the risks might be.
Amodei has warned that within 6–12 months, a swarm of AI agents could gain the ability to dominate the internet if protective measures remain optional. This scenario has intensified calls for independent audits, industry-wide benchmarks, and international coordination to address the rapidly escalating risks.
The AI debate echoes other moments when technology has moved faster than regulation. As reported earlier, gaps between innovation and real-world adaptation can leave communities exposed to unexpected consequences. The risks around AI now stand out as the most extreme example of this pattern.
What makes this moment different is the scale of the risk. Unlike earlier technological shifts, the possible downside of AI is not just economic disruption or privacy loss, but the survival of humanity itself. This has forced even optimistic developers to reconsider how quickly they are pushing the limits of what AI can do.
Despite the seriousness of Hubinger’s warning, there is little agreement on how to measure or reduce AI risk. Some argue that slowing development could block useful innovation, while others say the risks are too great to move forward without strong safeguards. What is clear is that the conversation has changed: existential risk is now central in AI policy, and the pressure to act is growing.
Given the speed of AI progress and the seriousness of the warnings from inside the industry, dismissing these concerns would be risky. When respected developers put the chance of human extinction at even 10%, it demands more than talk. The next few years will show whether the tech sector and policymakers are willing to put safety ahead of speed, or if the race for progress will continue until the risks are no longer theoretical.
