Confidential computing's trust mechanism is broken. The fix may not exist

New formal verification research demonstrates that current attested-TLS (intra-handshake attestation) designs can be abused in relay/diversion attacks that let clients unknowingly encrypt traffic to a malicious host; the flaw (CVE-2026-33697, score 7.5) affects multiple production implementations, was responsibly disclosed, and the authors recommend abandoning intra-handshake attestation in favor of post-handshake approaches that can cryptographically bind attestation to application traffic.