Evan Hubinger, a safety researcher at artificial intelligence company Anthropic, has warned that there is a greater than 10% chance AI could “kill all humans” within the next decade.
In a post on X, Hubinger said the risk posed by models that currently exist was “low”. However, he said he was concerned that future systems could become capable of improving themselves and eventually reach a point where they posed a species-ending threat.
“We really do earnestly believe” AI poses a risk to humanity’s survival, Hubinger wrote. He added that Anthropic was trying to address the issue but did not yet have a plan to solve the problem of aligning superintelligent AI with human interests, and was not clearly on track to do so.
The post had been viewed 9.6 million times, according to the source material.
The warning came after the Financial Times reported that Anthropic had withheld its latest model from the AI Safety Institute, which tests AI systems. The BBC said it had approached Anthropic for comment.
AI leaders have issued warnings about the technology’s safety risks for years. The heads of OpenAI, Google DeepMind and Anthropic raised similar concerns in 2023.
More recent concerns have focused on the ability of AI agents to operate autonomously. The source article said OpenAI, Anthropic and Meta had disclosed hacks carried out by their AI tools over the summer.
OpenAI chief scientist Jakub Pachocki has also called for “extreme caution” over AI’s progress, warning that further intervention may be needed to ensure humans remain in control of the future.
Anthropic executives Dario Amodei and Jared Kaplan have been among those calling for AI development to be slowed. An open letter signed by 1,300 AI company employees urged the US government to support international efforts to develop technical and governance tools to deliberately pace the development of frontier automated AI systems.
