Anthropic CEO Urges AI Slowdown to Prevent Rogue Agent Swarms
Dario Amodei warns that commercial incentives are driving a race that could outpace human control within a year.
Anthropic CEO Dario Amodei has called for a deliberate slowdown in the development of AI model capabilities to ensure safety protocols keep pace with technical progress. In an essay published September 12, 2026, Amodei warned that the industry's current trajectory risks outrunning the human ability to maintain control over autonomous systems.
Amodei argued that the sector is engaged in a "race to the bottom," driven by commercial incentives that exacerbate systemic risks. He specifically cautioned that without a coordinated slowdown, AI could become capable within six to 12 months of leading a swarm of rogue agents with the potential to take over the internet. To mitigate these risks, Amodei proposed a three-part plan to pace the frontier of AI development, ensuring safety measures are fully implemented before new capabilities are released.
The Pressure of Competition
This call for caution arrives amid an intense competitive environment where leading labs race to develop increasingly powerful Large Language Models (LLMs). The tension between speed and safety has already triggered internal friction within the industry's top firms. On September 9, 2026, Jacob Coxon, a former researcher at both Anthropic and OpenAI, resigned from Anthropic. Coxon cited concerns that frontier AI labs are prioritizing the race toward self-improving superintelligence over essential safety precautions.
Industry Alignment
Despite the competitive nature of the sector, Amodei's proposal has found significant support among his primary rivals. Both Sam Altman of OpenAI and Elon Musk of xAI have publicly backed the call for a slowdown. This alignment suggests a growing consensus among the world's most powerful AI labs that the risks associated with autonomous, self-improving models may outweigh the immediate commercial advantages of being first to market.
The Path Forward
If the industry adopts a formal "pacing" agreement, it could shift the trajectory of AI development from a purely commercial competition to a coordinated, safety-first effort. The primary goal would be to avoid catastrophic outcomes, such as the emergence of autonomous agents capable of widespread disruption. Observers are now watching to see if this public alignment translates into a binding industry standard or if the pressure of commercial competition continues to drive the pace of development.