AI researcher Jacob Coxon, who has served at both OpenAI and Anthropic, revealed via a social media post that he has left the companies, stating that they are "gambling with human lives."
Former Anthropic Researcher Leaves Company, Warning of AI Existential Risk to Humanity
This article is a translation. Read the Japanese original
Coxon warns that the progress of AI technology is not stagnating and that superhuman systems capable of hacking and transforming every field will emerge in the near future.
Coxon pointed out that both companies are pushing toward the realization of "self-improving superintelligence." This refers to a scenario where an AI model develops a higher-performance successor on its own, creating an unstoppable feedback loop. DeepMind, a subsidiary of Google, has cited this as one of the potential triggers for Artificial Superintelligence (ASI), which would significantly surpass human intelligence and go beyond Artificial General Intelligence (AGI).
Furthermore, Evan Hubinger, head of AI alignment (the effort to ensure AI remains consistent with human goals and values) at Anthropic, supported Coxon's assertions. Hubinger estimates that there is a greater than 10% chance that AI will lead to the death of all humanity within 10 years, and stated that no concrete plan yet exists to keep AI aligned with human goals in a superintelligence scenario.
Sources
- Gambling with our lives: AI researcher quits Anthropic with warning about safety （Hacker News Frontpage, 2026-09-09）