I routinely get rejected on elliptic curve related requests; often I can talk Claude back into some work on it. But it is surprisingly sensitive to talking about research cryptography in my experience. I’m not in their cyber program.