A draft executive order that would have pushed companies to submit advanced AI models for federal review and promoted voluntary safety assessments was obtained by the press but remained unsigned. The draft grew out of weeks of NDA-bound discussions with OpenAI, Anthropic and Google, and the new reporting shows those talks and internal divisions amplified concerns—like David Sacks's warning—that voluntary reviews might later be treated as mandatory. Its release follows an abrupt postponement amid White House infighting and industry worries about harming U.S. competitiveness with China.
Click a connection line between nodes to view confidence and evidence.