Artificial Intelligence (AI) could “wipe out all humanity” within the next decade. A top researcher at Anthropic, a company focused on AI safety, has warned that there is a greater than 10 percent chance of this happening.
Evan Hubinger, a researcher at Anthropic, stated in a post on the social media platform X that while the risk posed by currently existing models is “low,” the technology is advancing rapidly. He expressed concern that it could soon reach a stage where AI becomes capable of self-improvement, potentially posing an existential threat to humanity.
Hubinger did not provide a detailed explanation of how AI might wipe out humanity in the future. However, his comments represent the latest in a series of increasingly urgent warnings regarding AI.
The debate has largely shifted from whether AI is risky at all to a discussion about the extent of that risk.
Hubinger made these remarks in response to a post on X by Jack Clark, a former Anthropic researcher. According to a BBC report, Clark recently left Anthropic; he had previously worked at OpenAI.
Clark wrote that no organization is acting responsibly. He noted that these systems would soon evolve into entities with superhuman capabilities—capable of hacking anything, radically transforming any sector overnight, and acquiring real-world power and resources.
Computer scientist Dame Wendy Hall, who advises the United Nations on AI, expressed shock at the social media posts by Hubinger and Clark.
Speaking on BBC Radio 4’s World at One program, she suggested that some of these comments might be part of “public relations and marketing strategies,” given that both Anthropic and OpenAI are moving toward highly anticipated stock market listings. Meanwhile, the Financial Times has reported that Anthropic did not submit its latest AI model to the UK’s AI Safety Institute (AISI)—one of the world’s leading organizations for assessing AI risks.
Anthropic declined to comment on these posts or the situation regarding the AISI.
