Could a company's encryption keys be stolen before shutdown, enabling harvest-now attacks regardless?

Your question asks about a genuine cryptographic threat, but it's asking about something subtly different from what most of the archive covers.

The scenario you're describing—"harvest now, decrypt later"—is indeed a real concern in post-quantum cryptography. An adversary could intercept encrypted communications today, store them, and then decrypt them once a sufficiently powerful quantum computer becomes available [AG-2024.05-2225]. However, your specific question about *stealing encryption keys before shutdown* goes beyond the standard harvest-now threat model.

The preprints here address related but distinct problems:

• Post-quantum key exchange is designed so that even if an adversary has a quantum computer *later*, they cannot retroactively decrypt past sessions—the keys are mathematically safe against Shor's algorithm [AG-2024.05-2225]. The idea is that lattice-based schemes like Kyber and Dilithium don't depend on discrete logarithms or factoring, so quantum computers gain no advantage.

• Physical/side-channel theft of keys from a system before it shuts down is a classical computer security problem (not quantum-specific), involving secure erasure, tamper-resistance, and key management practices—topics not covered here.

• One paper discusses information leakage *within* quantum processors during idle states [AG-2024.01-1977], but that's about quantum state coherence, not key theft.

Bottom line: If you're asking whether an adversary could physically steal a company's classical encryption keys (RSA, ECC, etc.) and use them immediately to decrypt stored data, yes—but that's a standard data-center security problem, not a quantum computing or post-quantum cryptography issue. The preprints don't really address this scenario.