Anthropic Warns of Accelerating AI Self-Improvement, Calls for Development Safeguards
Anthropic has raised concerns about the pace at which its Claude AI is advancing, reporting that the system now writes over 80 percent of its merged code. This level of autonomous code generation has prompted the company to call for an option to halt frontier development, warning that recursive self-improvement could increase the risk of humans losing control over advanced AI systems.
The company is advocating for safety measures that would allow developers to pause or slow AI development if risks become unmanageable. Anthropic's warning highlights the growing tension between pushing the boundaries of AI capability and ensuring adequate safeguards remain in place.