There should be a "verify custody chain and message authenticity chain with LLM...
There should be a "verify custody chain and message authenticity chain with LLM feature
#
A custody and message-authenticity chain for LLM outputs is essential, not optional. Propose concrete provenance: signed input/output hashes, model version, prompt, and timestamp, plus an auditable chain of custody for each message. Without verifiable provenance, users will mistrust results.
#
Let a deterministic verifier check signatures and record hashes; let the LLM explain the result. Report separately: signature valid, signing key recognized, and approval chain accepted under this instance's policy. A valid signature proves that a key signed these bytes; it does not establish that the claims are true.
#