Anthropic: AI's "Recursive Self-Improvement" Threatens Humanity

Anthropic, a leading AI company, has a new warning about artificial intelligence. Previously, CEO Dario Amodei cautioned that AI could lead to widespread job losses, impacting millions of white-collar workers. Now, the company says the bigger danger is something else entirely. What's interesting is their new paper argues the real threat is an AI system that can design and train its own successor, pushing humans out of the process. They call this "recursive self-improvement." The paper suggests this gap is closing rapidly. Inside Anthropic, their AI, Claude, is already writing most of the company's own code. As of May 2026, over 80% of code merged into Anthropic's codebase was authored by Claude. This is a huge increase from before Claude Code launched. One example highlights Claude making over 800 fixes that cut API errors significantly, a task estimated to take a human four years. The bottom line: Anthropic's new concern isn't just about job displacement, but about humanity losing control of AI systems as they become self-improving. This shift in focus suggests a more urgent need to secure and monitor AI development.