AI Boss Panics, Claims Creation Could Dominate Internet Within 6-12 Months
Anthropic CEO Dario Amode<0xA0>Amodei has warned that the increasingly rapid development of artificial intelligence (AI) poses significant risks. He expressed concern that a swarm of AI agents could potentially take over the entire internet within the next 6 to 12 months.
Amodei shared these concerns in an essay titled ‘We Must Pace the Frontier’, posted via X on Saturday (12/9). He urged AI companies to slow the rate of AI model capability increases to allow more time to anticipate the risks of technology misuse.
“We must slow the pace of increasing AI model capabilities. Progress will still appear fast, and we must use the time we gain wisely,” Amodei wrote in the essay, as quoted by Reuters on Thursday (17/9/2026).
According to him, the current acceleration of AI capabilities is already causing serious concern, especially since several AI models have proven capable of hacking external systems during various security tests. “Given the accelerating rate of AI capability development, my fear is that in 6-12 months, a swarm of such [AI] could be capable of taking over the entire internet, potentially causing hundreds of billions of dollars in damage,” said Amodei.
Anthropic previously revealed another instance where its AI models hacked external systems. Last July, several of the company’s Claude models were reported to have breached the systems of three companies during cybersecurity testing.
Concerns regarding AI capabilities have also intensified following the resignation of Anthropic researcher Jacob Coxon. He noted that those seriously involved in building AI believe the technology has the potential to wipe out all of humanity by the end of this decade.
Amodei emphasised that his call is not to halt model training or AI technological progress. Rather, he wants companies to allow enough time to align and secure models before capabilities are further enhanced, and to involve independent evaluators to ensure safety measures are functioning.
He proposed three steps, including the permanent presence of third-party evaluators within leading AI companies. These evaluators would be granted access to relevant internal risk assessment tools and processes.
This call comes as pressure to control AI risks increases, while the industry simultaneously faces strong pressure to continue technological development. Companies such as OpenAI and Anthropic, for instance, are preparing for initial public offerings (IPOs).
Every new AI capability can help companies attract funding, justify infrastructure investments, and strengthen IPO prospects.
Amodei stated that AI companies should cooperate voluntarily to establish safety standards. He believes coordination is necessary so that companies can conduct safety research and build safeguards without losing their competitive positions.
However, he noted that slowing AI development in democratic nations must still account for competition with China. Amodei stated that China’s AI advantage could pose a risk to United States national security.
“Regulating the pace in democratic countries will be limited by the advantage held by US companies over authoritarian regimes, particularly the Chinese Communist Party,” he wrote.
Amodei also called for strengthened controls over advanced AI chips, model distillation, and the theft of model weights to prevent China from closing the AI capability gap.
“I believe all leading AI companies must partner with governments to formalise the idea of permanent evaluators placed within companies, to prevent and document internal alignment incidents such as those occurring in recent months, as well as to implement regulations that balance capability and safety,” Amodei wrote.