Modern cloud and enterprise systems face invalidation of identity-centric authorization as autonomous AI agents generate syntactically valid yet semantically unsafe actions, per the arXiv preprint 2605.15228. The Distributed Trust Framework (DTF) counters this via Justification Proofs encoding admissibility, consensus evaluation, ephemeral Execution Identities, and append-only Evidence Chains that enforce the invariant of no high-stakes execution without proof objects. This architecture is instantiated over OpenKedge-based governed mutation substrates and maps directly to cloud-native environments. DTF extends patterns from verifiable computation systems in prior work on secure enclaves (arXiv:2305.10978) and zero-trust policy engines (NIST SP 800-207). It shifts trust boundaries in sovereign AI deployments involving regulated data and national services by requiring distributed consensus for authority derivation. Evidence Chains enable replayable audits, addressing operational risks in financial workflows and infrastructure interactions under stated substrate assumptions.