diff --git a/docs/spec/threat-model.md b/docs/spec/threat-model.md
index 42d2385..922baa6 100644
--- a/docs/spec/threat-model.md
+++ b/docs/spec/threat-model.md
@@ -19,7 +19,14 @@ Status: Draft v0.1 | Covers: Phase 1 cMCP Runtime
 **A1: Rogue operator / privileged insider**
 - Has root access to the host VM
 - Can read/write host memory, modify files, access environment variables
-- Cannot read TEE-encrypted memory (SEV-SNP, TDX) or TPM-sealed secrets
+- On a memory-encrypting TEE (SEV-SNP, TDX): cannot read or modify guest memory while the
+  gateway runs, so policy evaluation and audit construction are outside this adversary's reach.
+- On the `tpm` tier: cannot unseal TPM-sealed secrets outside a matching measured boot, and
+  cannot alter the measured components without the attestation reporting it. This is a
+  boot-time and at-rest guarantee only. **A TPM does not encrypt the memory of a running
+  process**, so on bare-metal TPM an operator with root can read or modify the gateway's
+  memory while it runs, including flipping a Cedar decision, without the TPM detecting it.
+  The `tpm` tier defends against a tampered boot, not against a live privileged insider.
 - Goal: bypass policy, forge audit entries, or exfiltrate data
 
 **A2: Supply chain attacker**