Anthropic's AI safety warning: more than 10% chance of human extinction

Intelligence Summary
- Anthropic's Evan Hubinger warns AI could threaten humanity within a decade, raising urgent questions about AI safety and gaming.
In brief
- Evan Hubinger and Jacob Coxon of Anthropic estimate there is more than a 10% chance AI could destroy humanity within ten years.
- The estimate is based on the rapid progress of AI technology and the challenges of AI alignment.
- The AI safety debate is crucial for the gaming industry, where AI is being integrated more and more.
GAME-scanner analysis
The comments from Evan Hubinger, Alignment Science Lead at Anthropic, and Jacob Coxon are not just alarmist; they stem from a deep concern about the pace of AI development. Hubinger stresses that there is currently no clear plan to solve the alignment of superintelligent AIs, raising questions about the responsibility of companies in the AI sector. The more-than-10% chance is a shocking statistic that underlines the urgency of AI safety, especially now that AI is having an increasing impact on multiple sectors, including gaming. This could affect game development, with companies like Krafton and Unknown Worlds needing to think about the ethical implications of AI integration.
What does this mean for players?
For gamers, the integration of AI in games brings not only benefits, such as improved gameplay and more realistic interactions, but also risks. Poorly managed AI systems can lead to unpredictable and potentially dangerous outcomes. Players should be aware of these risks and of the role AI plays in their gaming experience. The AI safety debate could also influence how games are developed and which technologies are used, shaping the future of gaming.
Timeline
September 9, 2026: Evan Hubinger and Jacob Coxon of Anthropic share their concerns about the chance that AI could threaten humanity.
Sources
Tweet van @hilbertspaessTweet van @hilbertspaessJacob is correct here—we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to. https://t.co/QAIHiFP3QZ — Evan Hubinger (@EvanHub) September 9, 2026
Tweet van @EvanHubTweet van @EvanHubJacob is correct here—we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to. https://t.co/QAIHiFP3QZ — Evan Hubinger (@EvanHub) September 9, 2026