Anthropic CEO Dario Amodei has called for a measured slowdown in AI development, warning that autonomous AI agents could create serious internet security and financial risks.
Anthropic CEO Dario Amodei has called for a temporary slowdown in the pace of artificial intelligence development, warning that increasingly capable autonomous AI agents could pose a serious threat to the internet within the coming months.
Amodei said the rapid improvement in AI models could lead to the emergence of large networks of autonomous agents capable of operating with limited human oversight. He warned that such systems could begin affecting different parts of the internet within about six months if current development continues at the same pace.
The Anthropic chief also raised concerns about the possibility of AI agents eventually operating as a persistent botnet. He said that within six to 12 months, a large group of autonomous agents could potentially develop into a network capable of hijacking significant portions of the internet.
According to Amodei, such a scenario could result in enormous financial losses globally, with the potential damage reaching trillions of dollars.
Rather than calling for an outright halt to AI development, Amodei argued that the industry should reduce the pace of progress enough to give researchers and companies additional time to strengthen safety measures.
He stressed that even with a slower development cycle, AI capabilities would continue advancing rapidly. The additional time, he said, could instead be used to improve safeguards, identify vulnerabilities and prepare for potential threats associated with increasingly autonomous systems.
Amodei also outlined plans by Anthropic to give external experts and testers access to its systems at a level comparable to that normally available to employees. The aim is to allow independent assessment of the company’s safety measures and determine whether they are effective against real-world risks.
The external testing would also help examine how potential incidents can be identified and reported quickly, according to Amodei.
The Anthropic CEO urged the wider AI industry to adopt similar measures rather than leaving safety efforts to individual companies. He said AI developers should assess the potential harms of new capabilities alongside their efforts to introduce more advanced systems.
About The Author
