Anthropic says AI could build its own successor, calls for slowdown in development

Anthropic, the AI company behind Claude, has urged leading firms to slow the development of advanced AI systems due to rapid progress toward autonomous self-improvement. In a blog post, the company stated that AI could eventually design and build its own successors without human intervention, a process called recursive self-improvement. While not yet achieved, Anthropic’s internal data shows Claude is already accelerating AI development by assisting engineers in writing code, increasing productivity and handling more complex tasks. The company emphasized that unchecked advancements could lead to alignment issues and loss of control, despite potential societal benefits in medicine, technology, and the economy. Anthropic proposed establishing a global framework to enable temporary pauses in AI development if advancements outpace safety measures. It also suggested creating a verification system to ensure compliance with any agreed-upon slowdowns. On June 4, 2026, Anthropic shared on X (formerly Twitter) that its internal data reveals Claude’s role in accelerating AI development, raising concerns about the speed of progress. The company plans to host discussions in the coming months to address questions about recursive self-improvement and its implications. Anthropic’s call follows observations that AI systems are evolving faster than institutions can prepare for, highlighting the need for coordinated action to mitigate risks.