Memory Provenance Laundering: How LLM Agents Lose Trust in Long-Term Memory

richie9830 · reddit · 2026-08-10

This paper explores the issue of "Memory Provenance Laundering" in LLM agents. During memory consolidation, untrusted external observations can be transformed into what appears to be trusted user history or workflow context.

The original source and its trust level often disappear during this process, but the action trigger remains. As a result, subsequent agents inherit the conclusion without inheriting why it should be trusted. The paper suggests preserving provenance through memory consolidation and matching the authority of a memory to the risk level of the action it enables.

Original post →

More from coding & agent

coding & agent channel →