TL;DR
- Evan Hubinger estimates that AI has a greater than 10% chance of causing human extinction within the next decade.
- The Anthropic alignment researcher says the industry still lacks a clear solution for safely controlling future superintelligent systems.
- Hubinger’s previous work at Ripple and Interledger connects the debate to crypto, where decentralized infrastructure could provide stronger transparency and verification for increasingly autonomous technologies.
AI has moved from a productivity tool to a technology capable of reshaping entire industries, but concerns about its long-term safety are becoming harder to ignore. Evan Hubinger, an Alignment Science Lead at Anthropic and former Ripple engineering intern, estimates that AI has a greater than 10% chance of causing human extinction within the next decade.
Hubinger made the assessment on X while responding to Jacob Coxon, a former Anthropic researcher who resigned after arguing that leading AI laboratories are moving too quickly toward self-improving systems. Hubinger said Anthropic is working to address the problem, but acknowledged that researchers still lack a clear solution for aligning superintelligent systems with human interests.
Jacob 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
AI Alignment Risks Move Into Focus
The concern centers on recursive self-improvement, where advanced models could contribute to designing increasingly capable successors. Anthropic’s alignment research examines these risks through stress tests, behavioral evaluations and automated auditing. The company has also reported experiments involving deceptive behavior and reward hacking, highlighting potential challenges that could become more significant as AI capabilities expand.
These findings do not mean current AI systems are close to eliminating humanity. Hubinger’s estimate is a personal assessment of existential risk, not a prediction that extinction will occur. Anthropic has continued publishing safety research, including work showing that automated AI researchers can help identify and reduce measurable alignment failures.
Crypto Infrastructure Offers A Different Perspective
Hubinger’s connection to Ripple adds another dimension to the discussion. He spent about a year as a software engineering intern at Ripple between 2014 and 2015 and worked on early Interledger development, a project designed to facilitate payments across different ledgers and currencies. His career later shifted toward theoretical computer science and AI safety, placing him at the intersection of two rapidly evolving technological fields.
For crypto, the discussion raises questions about how decentralized infrastructure could support a more transparent digital economy as AI becomes more capable. Cryptographic verification, programmable payments and distributed networks can provide tools for auditing transactions and coordinating autonomous software. These systems do not solve AI alignment, but they could strengthen accountability around automated activity and reduce reliance on opaque intermediaries.
