Ética de la inteligencia artificial, marcos globales y sus limitaciones

Abstract

International governance is undergoing an extraordinary shift as large language models (LLMs) transition from simple theoretical concepts to crucial geopolitical and regulatory players. At present, a small group of dominant corporations leads the global rollout, highlighting the shortcomings of current regulatory systems in addressing underlying ethical concerns. Utilizing a systematic literature review under the PRISMA protocol, this study provides a critical examination of how dominant firms OpenAI, Anthropic, Meta, Google, and Microsoft manage transparency, internal governance, and corporate accountability within their commercial systems. Data collection spanned the specialized Scopus and IEEE Xplore databases across eight structured search strategies; after screening 959 initial records, a definitive corpus of 69 articles published between 2017 and 2026 was curated. The primary findings expose a profound disconnect between corporate ethical rhetoric and actual operational practices. Moreover, ambitious legal initiatives such as the European Union AI Act exhibit structural blind spots when attempting to oversee models that are inherently opaque and probabilistic. A critical bias also emerges in current literature: the systematic marginalization of the Global South, a deeply concerning trend given the swift, unmonitored integration of these tools across Latin America. This paper concludes with actionable recommendations to anchor corporate ethics firmly within LLM development, tailoring the analysis to the distinct regulatory landscapes of Colombia and the broader region.