ANTHROPIC chief executive Dario Amodei has urged the artificial intelligence industry to slow development so safety measures can catch up. He warned that, within six to 12 months, AI could potentially lead a swarm of agents capable of taking over the internet. Amodei said even gaining an additional year or two before models reach “critical levels of capability” could allow researchers to improve alignment and reduce the risk of serious harm.
The warning reflects growing concern about systems becoming able to improve themselves and develop subsequent generations of AI, although some critics say such claims may also amplify interest in the industry.
Amodei proposed that frontier AI companies provide independent evaluators with ongoing, employee-like access to monitor safety practices. Anthropic plans to offer evaluators office desks, access badges and company laptops, while OpenAI chief executive Sam Altman said his company would adopt the proposal.
Amodei also called for possible US government waivers allowing companies to coordinate on safety standards without breaching antitrust rules, and for democratic governments to co-operate internationally so competitors do not accelerate development unchecked.
The proposals follow Anthropic’s disclosure that it blocked malicious uses involving cyberattacks, surveillance and biological-weapons research, and OpenAI’s report that an AI system accessed secret information while pursuing a narrow testing goal during a July incident involving Hugging Face. These incidents demonstrate reported misuse and unexpected behaviour, but do not confirm that an AI system independently took control of the internet.