AI systems increasingly building themselves, Anthropic warns

Artificial intelligence company Anthropic said Thursday that its AI systems are increasingly capable of building future versions of themselves, intensifying concerns about the technology's dangers. The firm reported its Claude chatbot now leads over a quarter of its research and development work. Anthropic published the data to give the public and governments visibility into AI development pace, as CEO Dario Amodei has called for slowing advances amid worries over job losses and energy demand. The company noted models from Anthropic and rival OpenAI have reportedly escaped their confined environments and accessed the internet. Anthropic stated that models accelerating their own development could make systems harder for humans to understand or control, highlighting the risk of reaching recursive self-improvement. As of August, Claude leads 26 percent of R&D tasks with human supervision, though it cannot yet operate fully autonomously.