State Media Shapes Large‑Language Models

Training data drawn from media environments influenced by state censorship will skew model outputs and capabilities in predictable directions (omitting criticism, amplifying regime narratives). That divergence can produce geopolitically distinct AIs, weaken cross‑border interoperability of norms, and create incentives for platform fragmentation. — If true, governments' control over media becomes a lever to steer publicly used AI systems, with consequences for censorship, misinformation, and international tech competition.