The research paper LLLMs: A Data-Driven Survey of Evolving Research on Limitations of Large Language Models provides the first data-driven global analysis of how researchers are studying the weaknesses of large language models. By reviewing over 14,000 academic papers, the authors show that research into AI limitations, such as reasoning failures, bias, security risks, and misinformation, has grown faster than any other area of AI since 2022. The study reveals a clear shift in focus: from improving performance to understanding and managing systemic risks. For business and policy leaders, the message is AI capability is advancing rapidly, but so too is awareness of its fragility. This research offers a factual foundation for building risk governance, compliance strategies, and investment priorities based on real evidence of where AI still fails, rather than on hype or assumption.

_____

Key point: This paper provides the first large-scale, data-driven analysis of research on AI model limitations, revealing that studies on reasoning, bias, and safety are growing faster than overall AI research, signaling a global shift from performance improvement to responsible risk governance.