David Sacks argues that OpenAI and Anthropic can prioritize safety without waiting for government action. The harder question is whether voluntary restraint and existing laws are enough.