Anthropic Researcher Warns AI Could Kill Us All After Quitting | CoinDesk Daily

Jennifer Sanasie (0:00)

One anthropic researcher quit their job and warns AI could kill us all, and Iran turns to crypto to dodge US sanctions. This is CoinDesk Daily.

Want trusted crypto research? Check out The Stack, Grayscale's blog. Visit grayscale.com/thestack.

RealFi, a smarter stablecoin backed by real world assets. Join the testnet now at realfi.co. One Anthropic researcher quit their job with a warning that the AI labs are racing towards self-improving super intelligence and quote gambling with our lives. He says the people building AI earnestly believe it could kill everyone by the end of the decade. Anthropics own alignment lead, Evan Hubeinger, backed him up, putting the odds above 10% within the next decade. Saying the AI world US security agencies are accusing six Chinese AI companies, including DeepSeek, Alibaba and Moonshot of using American frontier models to train their own systems. The FBI, NSA and CISA say the firms extracted billions of tokens from models like Claw, GPT and Gemini since late 2024, using proxy services and multiple accounts to avoid detection. They're calling it, quote, malicious knowledge distillation, likely done with Beijing's awareness. And Iran's central bank is quietly encouraging traders to bring overseas earnings home using crypto to keep trade flowing under tightening US sanctions. Authorities have eased foreign currency controls, letting exporters settle cross-border deals through Iranian crypto exchanges at market rates. Previously, they had to route earnings through government platform at below market rates. Iran estimates businesses are sitting on more than $100 billion in undeclared earnings. Get more updates on coindesk.com. Want trusted crypto research? Check out The Stack, Grayscale's blog. Visit grayscale.com/thestack.

RealFi, a smarter stablecoin backed by real world assets. Join the testnet now at realfi.co.