OpenAI diverges from Trump's AI EO in a new policy paper, proposing cyber risk evaluations for advanced AI systems be mandatory and led by CAISI, not the NSA

OpenAI is charting its own course on AI safety governance — and it doesn't align with the White House.

In a new policy paper, the company proposes that cyber risk evaluations for advanced AI systems should be mandatory. The twist: OpenAI wants those evaluations led by CAISI, not the NSA as outlined in Trump's AI executive order.

That's a notable split from the administration's preferred approach to AI oversight. The NSA has traditionally handled signals intelligence and cybersecurity at the national level. OpenAI apparently thinks a different body should call the shots on frontier AI risk assessment.

The timing is deliberate. CEO Sam Altman is in Washington this week for a series of Wednesday meetings with White House officials, making the paper a clear lobbying play ahead of face-to-face conversations with the people who actually set policy.