As the United States and China prepare to negotiate artificial intelligence safety, a deeper structural problem shadows the table: governments can make commitments, but they cannot always compel the companies beneath them to honor those commitments. A former US AI diplomat warns that Beijing's stated caution about AI risks has not reached the domestic firms actually building the technology, creating a credibility gap that diplomatic language alone cannot bridge. History suggests that agreements between states are only as meaningful as the enforcement mechanisms and corporate cultures that give