Perfect open-source infrastructure for AI is necessary for safety because proprietary systems concentrated in a few Western companies would allow those companies to control human opinion and culture, creating regulatory capture, while no other government would accept American companies controlling super-intelligent AI systems.