Jack Paul 
Evan Hubinger, a former Ripple software engineering intern and now an Anthropic researcher, says the odds of artificial intelligence causing human extinction are higher than many people assume. He puts the chance at more than 10% within the next decade. The estimate came as he responded to Jacob Coxon, a former Anthropic researcher who resigned on Wednesday.
Coxon left with a public warning. He accused leading AI labs of racing toward self-improving superintelligence without enough safety measures. In his view, neither OpenAI nor Anthropic is acting responsibly, and he said they are gambling with human lives. He also argued that frontier systems may soon gain real influence outside controlled settings. Such systems, he warned, could hack almost anything, reshape entire fields overnight, and acquire power and resources.
Hubinger agreed with the core concern. On X, he wrote that Coxon was correct and that Anthropic researchers believe AI could kill all humans. He added that he personally thinks the risk is above 10% over the next decade. But he did not frame this as proof that Anthropic is doing nothing. Hubinger said he believes the company is trying its best, while adding that no plan yet solves alignment for superintelligence. That distinction matters. Hubinger works on alignment stress-testing at Anthropic, which means he studies how safety techniques might fail. His warning draws attention partly because he is directly involved in the work he is criticizing.
OpenAI and Anthropic releases add urgency
The comments landed just days after OpenAI and Anthropic released some of their most capable models. OpenAI introduced GPT-6 Astra on Sept. 3. Nvidia CEO Jensen Huang described it as the arrival of superintelligence. Anthropic, meanwhile, launched Claude Fable 5.1 and Claude Mythos 5.1, which the company calls its most advanced models for coding and scientific research. The timing has made the safety debate feel more immediate, even if the exact risks remain hard to measure.
Coxon said major labs are not properly disclosing AI-related risks to the public. Still, he said he remains optimistic that they could coordinate to slow the race. That may sound contradictory: warn about an emergency while holding out hope for cooperation. But it reflects a split many researchers feel. They want labs to move faster on safety without pretending the commercial race will pause on its own.
From Ripple to AI safety
Before AI alignment became his focus, Hubinger spent about a year as a software engineering intern at Ripple, from June 2014 to August 2015. He worked on the early design of Interledger, a protocol meant to move payments across different ledgers and currencies. He later moved into theoretical computer science and AI safety, including work at the Machine Intelligence Research Institute.
Ripple CEO Brad Garlinghouse has taken a different public stance on AI. He has said Ripple is adopting AI aggressively and predicted it could boost growth. Garlinghouse also admitted his optimism is somewhat contrarian. Hubinger’s warning puts a darker estimate into a debate that is already moving quickly.