Law and Economics of Language Model Development: Empirical Examination of Corporate Strategies and Vaporware Claims

Abstract

Abstract This research investigates the application of corporate strategies in developing Large Language Models (LLMs) like ChatGPT, with a focus on law and economic aspects. Improved LLM performance is largely credited to expanded dataset size, leading to developments of similar models in non-English languages. This study questions whether such announcements, particularly in the Japanese market, could be classified as ‘vaporware’ posing potential antitrust issues. Using a stock event approach, the research scrutinizes the possibility of vaporware characteristics in these announcements by examining any resultant cumulative abnormal returns (CARs). The empirical evidence suggests an absence of significant upsurges in CARs in response to these announcements, implying a lack of vaporware characteristics and a stable market response. Consequently, in a market displaying reasonable efficiency towards investment in LLM development, these findings underscore the necessity for meticulous contemplation of competition policy regulation and the implementation of industrial policy promotion measures.