Cuban official discourse disproportionately present in AI training corpora via non-Cuban channels

Analysis finds that official Cuban state discourse is heavily represented in AI training datasets, but almost exclusively through non-Cuban sources, indicating a potential bias in how Cuban narratives are captured and propagated by AI systems. The finding raises questions about the authenticity and framing of Cuban content in global AI models, though the methodology and scope of the analysis are not detailed.