Around The IT World

Professional Tech News from 23+ Languages

🇷🇺 Russian, 6 November 2026, self-improvement

Anthropic Warns of AI's Approach to Recursive Self-Improvement

The company, developer of the Claude AI, warns of the potential for AI to significantly accelerate its own development.

Anthropic, the developer of the Claude AI, has warned of the potential for AI to significantly accelerate its own development, a phenomenon known as recursive self-improvement. This could lead to a self-sustaining cycle of technological progress. According to Anthropic’s policy on responsible development, one of the criteria is when AI can achieve a year’s worth of progress in just one year, without human assistance. However, none of the current models of Claude have officially reached this level.

The company has already seen AI actively participating in the development of its own successors, with Claude writing over 90% of the code added to the company’s projects in September 2026. The AI also helps with experiments, data processing, and research results verification.

The main concern of the developers is that security mechanisms may not be able to keep up with the growing capabilities of AI. Anthropic has acknowledged that it has not always been able to timely detect the acceleration of its own research. Additionally, errors made by one model when creating the next may be passed on to new generations of neural networks.

Back to the latest news