AI TRENDS | Anthropic Warns Recursive Self-Improvement in AI Could Pose Major Risks

Anthropic has warned that recursive self-improvement (RSI) in artificial intelligence could bring significant risks. According to Odaily, the company said last week that its AI model Claude can now generate about 80% of the company’s code. Anthropic described this as a key signal that AI is entering a “self-iteration” stage, in which a model can design the next generation of AI systems without human intervention. The concept of RSI has recently drawn broad attention in Silicon Valley. Last month, OpenAI held a related conference in San Francisco that brought together multiple researchers to discuss the potential and risks of AI self-iteration. Anthropic said that as AI capabilities improve rapidly, monitoring for potentially uncontrollable behavior and building safety mechanisms are becoming increasingly urgent.