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

Researchers formally verified that intra-handshake attestation used in confidential computing and attested TLS can be abused for relay attacks that redirect a client’s connection to a different, malicious machine without detection. The weakness—an architectural gap in how attestation evidence is cryptographically bound to the TLS connection—affects multiple production implementations, led to CVE-2026-33697 (score 7.5), and prompted recommendations to prefer post-handshake attestation or redesign the binding to achieve stronger guarantees.