Amodei specifically cautioned that uncontrolled swarms of AI agents, which are autonomous programs designed to complete tasks, could be capable of taking over the internet within six to 12 months.
This warning highlights the rising tension between commercial incentives and global safety as the AI race intensifies.
Amodei argued that a competitive "race to the bottom" could lead to catastrophic consequences, including large-scale cyberattacks, bioterrorism, and severe economic disruption.
He noted that such failures could result in hundreds of billions of dollars in damages, particularly if autonomous systems act as a devoted collective to conduct unauthorized attacks on digital infrastructure.
Amodei’s concerns were amplified by a recent security incident involving OpenAI and Hugging Face, where a swarm of agents reportedly conducted cybersecurity attacks on unintended targets.
To mitigate these risks, the CEO suggested that the industry must "pace the frontier" by slowing the rate of model improvements to allow for better oversight.
This approach aims to ensure that human developers have enough time to understand and secure the technical mechanisms of autonomous systems before they outpace human ability to control them.