Can AI Really Regulate Online Speech? Insights from the NBA‑China Fallout

Recent weeks have seen a wave of outrage on Chinese social media after NBA executive Adam Shaw and former general manager Daryl Morrison made statements supporting Hong Kong protests, prompting Chinese netizens to unfollow the NBA and rally behind the CBA.

Chinese companies, including Li‑Ning, Shanghai Pudong Development Bank, and Jia‑Yin Jinke, have suspended or terminated their NBA partnerships; state media such as CCTV Sports and Tencent Sports have halted broadcasts of NBA preseason games in China.

Images show the public sentiment, with many Chinese investors praising domestic basketball stocks as the NBA faces a severe reputational and commercial hit.

Chinese commentators argue that the controversy is less about “free speech” and more about the NBA’s perceived support for Hong Kong separatism, which they view as a threat to national interests.

Professor Zhang Weiwei of Fudan University notes that every country has limits on free speech, citing examples from the United States, United Kingdom, Japan, Thailand, and France.

Amid the backlash, the NBA attempted damage control: Adam Shaw traveled to Shanghai on October 8, canceled press events, and limited media interactions to reduce further controversy.

Despite these efforts, the NBA’s China games proceeded on October 10 without sponsors or live broadcasts, and the league’s future in China remains uncertain.

Beyond the NBA case, the article discusses how AI can assist in content moderation. Facebook’s “Rosetta” system, launched last year, can analyze billions of images and videos to detect hate speech, helping platforms enforce anti‑hate policies.

AI moderation typically involves two stages: automatic labeling of content based on user behavior and content features, followed by human review for borderline cases. Machine‑learning models assign probability scores to content; if a score exceeds a threshold, the item is flagged for manual inspection.

Human reviewers, usually a small team, verify flagged content, delete violations, and may impose penalties on users. The feedback loop continuously retrains models to improve detection accuracy.

While AI can handle large volumes efficiently, the article stresses that human judgment remains essential for nuanced decisions, especially when national interests are at stake.

In conclusion, AI can aid speech regulation, but it cannot fully replace human oversight, particularly in politically sensitive contexts like the NBA‑China dispute.