Arabic version: باحث في أنثروبيك يحذّر من خطر محتمل لانقراض البشر بسبب الذكاء الاصطناعي

According to BBC News, Anthropic safety researcher Evan Hubinger said he believes there is a greater than 10% chance that artificial intelligence could kill all humans within the next decade. He wrote on X that the risk from models currently available was low, but said he was worried AI could soon develop and improve itself to a point where it posed an existential risk.

Hubinger did not explain how future AI systems might lead to humanity being wiped out. His comments came in response to a post by Jacob Coxon, an AI researcher who said he had quit Anthropic and previously worked at OpenAI. Coxon wrote that neither company was acting responsibly and said systems would soon become superhuman, able to hack anything, revolutionise fields overnight and acquire power and resources.

Hubinger, whose post has been viewed more than 10 million times, said Anthropic was trying its best but did not yet have a plan to solve alignment for superintelligence. AI alignment seeks to build human ethical ideas and principles into the technology, keeping systems on track with what people value.

Separately, the Financial Times reported that Anthropic withheld its latest model from the UK’s AI Safety Institute. The BBC approached Anthropic for comment. A Cabinet Office spokesperson did not say whether the model had been withheld, but said it continued to collaborate closely with industry partners, including Anthropic, to make models safer.

Anthropic’s August safety report said there was a low risk of highly capable AI carrying out automated research and development that could cause catastrophic harm initiated by the AI, though it said it was less confident in that assessment than previously. OpenAI, Anthropic and Meta have also disclosed cyber-attacks carried out by their AI tools, while leading figures have called for greater caution and a slower pace of frontier AI development.