Anthropic researcher resigns with warning about the dangers of AI development, tell me about this news in (https://www.wral.com/news/ap/2ed54-anthropic-researcher-resigns-with-warning-about-the-dangers-of-ai-development/)

The news reports that an AI pretraining researcher, Jacob Coxon, has resigned from Anthropic while issuing stark public warnings regarding the rapid and irresponsible pace of artificial intelligence development across the industry.Here are the key details highlighted in the reports:

1. The Core Warning: Coxon, who previously worked on model pretraining at OpenAI before joining Anthropic, stated that leading AI laboratories are recklessly racing toward “self-improving superintelligence” and effectively gambling with humanity’s future. He warned that upcoming systems will soon be capable of autonomous hacking, acquiring real-world power and resources, and causing dramatic disruptions overnight.
2. Claims Regarding Industry Mindset: He asserted that many people building advanced AI systems earnestly believe the technology could pose an existential threat to human life within the next decade. While noting that Anthropic is more aware of these stakes than rivals, he argued that competitive pressures force them into the race anyway under the assumption that they must get there first to ensure safety.
3. Anthropic Internal Reaction: Following Coxon’s statements, Anthropic’s Alignment Science Lead, Evan Hubinger, publicly concurred with concerns about existential risk, estimating the likelihood of catastrophic outcomes within the next decade at greater than 10%, and noting that the industry does not yet have a verified alignment solution for recursively self-improving superintelligence.
4. Broader Context: The resignation comes during intense market competition and public preparations for initial public offerings, alongside escalating discussions among US lawmakers and global regulators regarding whether mandatory slowdowns, moratoria, or tighter oversight standards should be imposed on frontier model training.