Anthropic: Slow Down Frontier AI for Society to Catch Up

Anthropic suggests that artificial intelligence is advancing so rapidly, leading AI labs might need to slow down. The company's research arm, The Anthropic Institute, states that AI is already accelerating the development of new AI models. This could eventually lead to AI building its own successors. Anthropic believes a pause or slowdown in frontier AI development would allow societal structures and alignment research to catch up. The company's internal data shows significant delegation to AI, with over 80% of code merged into its codebase now written by Claude. One employee noted it had been roughly five months since they last wrote any code themselves. While current AI models still struggle with higher-level judgment, their capabilities are rapidly advancing. Anthropic warns that any meaningful slowdown would require coordination among multiple developers and governments, as a unilateral pause would do little to improve safety. Building the necessary infrastructure and trust for such coordination takes time, and Anthropic states, "We don't have that long." This highlights a critical challenge for managing the future of AI.