Confidential computing, promoted as a foundation for Europe's sovereign cloud strategy, is undermined by a cryptographic protocol flaw that may not have a solution. The technology relies on remote attestation to prove a Trusted Execution Environment runs genuine software, and vendors claim full control over customer data. Research by Muhammad Usama Sardar of TU Dresden and co‑authors showed that the attested TLS protocol can be subjected to relay attacks, allowing a client to verify a genuine server while its traffic is encrypted to a compromised machine. Their paper, Identity Crisis in Confidential Computing, presented at AsiaCCS 2026, demonstrated that connection evidence can be silently redirected to a different server anywhere in the world. A follow‑up study examined seven intra‑handshake binding mechanisms and found none prevent relay attacks. The analysis covered four production systems—Meta's Private Processing for WhatsApp, Contrast, Cocos AI, and a CCC proof‑of‑concept—all vulnerable to the flaw. The vulnerability was assigned CVE‑2026‑33697 with a severity rating of 7.5. The IETF TLS working group acknowledged the attacks, while the CCC's Attestation Special Interest Group delayed publishing the proof for more than a week despite repeated requests, leading Sardar to release the artefacts independently. Germany's Federal Office for Information Security (BSI) stated that confidential computing is a defense‑in‑depth component that does not guarantee key or identity management, and that vendor claims about its ability to ensure digital sovereignty are overstated. Intel said its attestation infrastructure does not limit sovereignty guarantees and does not receive plaintext data, while Google did not respond to inquiries. Sardar recommends abandoning intra‑handshake attestation in favor of post‑handshake attestation, noting that the timing of evidence binding makes level three security unattainable under the current architecture.