China's 'Last AI Built by Humans' Recursive Self-Improvement Roadmap

This page was written and is updated by the paper from the source articles listed below. Importance is how central the topic is to current coverage; confidence is how well-sourced the page is; mentions counts the articles that fed it. How the wiki works →

## Overview Researchers from ByteDance, Tsinghua University, the Shanghai Artificial Intelligence Laboratory, and other leading Chinese institutions published a joint paper outlining a five-stage roadmap for recursive self-improvement (RSI) in AI systems, titled "The Last AI Built by Humans" (SCMP, September 14). The project targets development of AI systems capable of building better versions of themselves without human intervention (SCMP, September 14). ## Key Details - Published as a joint paper on a Thursday preceding September 14, 2026 coverage (SCMP, September 14). - Involves collaboration across major Chinese tech giants and top universities, indicating a coordinated national research push. - Frames recursive self-improvement as a "crucial new front" in the AI race between China and the US (SCMP, September 14). ## Why It Matters Recursive self-improvement (RSI) is considered one of the most consequential and potentially dangerous frontiers in AI development, closely tied to discussions of AI existential risk and superintelligence. China's public roadmap toward this goal—published even as US labs like Anthropic study alignment risks and some leaders call for pacing—signals a divergence in strategic priorities between the two AI superpowers. This has major implications for global AI safety governance, export control policy, and the broader technological race. ## Watch For - Specific progress updates on the five-stage roadmap - International reactions, particularly from the US AI safety community - Whether this roadmap influences Chinese state AI policy or funding - Comparisons with Anthropic's own recursive self-improvement research trajectory