Play store
Artificial intelligence could advance so rapidly that it may have the potential to kill all humans within the next decade, according to Evan Hubinger, a senior safety researcher at AI company Anthropic.
Hubinger said in a personal assessment that he believes there is more than a 10% chance of such a catastrophic outcome.
According to a BBC report published Wednesday, Hubinger raised the concern in a post on X. He said existing AI models pose relatively low risks, but warned that the technology could soon reach a stage where AI systems are capable of improving their own capabilities.
Such a development could create a major threat to the survival of humanity, he said.
Hubinger stressed that Anthropic is making every effort to ensure the safety of its AI systems. However, he said there is still no effective solution or clear plan for keeping superintelligent AI under human control.
The comments came amid a report by the Financial Times that Anthropic had not provided its latest AI model to the UK’s AI Safety Institute, one of the world’s key organisations involved in assessing AI risks.
The BBC said it contacted Anthropic for comment but did not immediately receive a response.
The UK Cabinet Office also declined to comment on whether the latest model had been provided to the AI Safety Institute. It said, however, that the government is working closely with industry partners, including Anthropic, to make AI models safer.
Neil Lawrence, professor of machine learning at the University of Cambridge, told BBC Radio 4 that the Financial Times report appeared credible. He also suggested that competition between the United States and China over AI development, along with Washington’s increasingly isolationist stance, could lead to reduced technological cooperation with allies.
Concerns over AI safety are not new. In 2023, senior figures from OpenAI, Google DeepMind and Anthropic warned about the potential risks posed by the technology.
More recently, concerns have grown over AI agents capable of operating autonomously. In recent months, OpenAI, Anthropic and Meta have disclosed cases involving cyberattacks carried out with the assistance of their AI tools.
OpenAI chief scientist Jakub Pachocki has also called for “extreme caution” as AI capabilities continue to advance.
Calls to slow the pace of AI development are also gaining momentum. Several prominent technology figures, including Anthropic executives Dario Amodei and Jared Kaplan, have supported slowing AI development.
In an open letter signed by around 1,300 employees of AI companies, workers also called for an international effort to deliberately control the pace of development of advanced AI systems.
Source: BBC
