Anthropic researcher believes more than 10% chance AI 'could kill all humans'

According to the Saudi Gazette, a top safety researcher at the AI firm Anthropic has publicly warned that there is a greater than 10% chance that artificial intelligence could pose an existential threat to humanity within the next decade. The comments were made by Evan Hubinger in a post on the social media platform X, formerly known as Twitter. While Hubinger noted that the risks from current AI models remain low, he expressed significant concern regarding the rapid pace at which the technology is evolving and its potential to eventually improve itself beyond human control.

Hubinger’s remarks followed a separate announcement from Jacob Coxon, another AI researcher who recently resigned from Anthropic. Coxon, who previously held a position at OpenAI, stated in his own post that he was leaving the company due to concerns that AI laboratories are taking excessive risks with public safety. While neither researcher provided a detailed technical roadmap for how such an existential threat might manifest, their statements highlight a growing divide within the technology sector regarding the speed of AI development and the adequacy of current safety guardrails.

This exchange is part of a broader, intensifying debate among industry experts and researchers about the long-term implications of advanced artificial intelligence. As the discourse shifts from questioning whether AI poses a risk to determining the scale and probability of that risk, these public interventions from former and current employees serve as a barometer for the internal pressures facing major AI labs. Anthropic has not yet issued a formal response to the specific claims made by the departing researchers.