AI & Alignment
Alignment After Witnessing: Governance at the Edge of Intelligibility
In production (complete)A boundary analysis of what alignment can still mean once witnessing has already degraded or disappeared.
Function in corpus
Edge-case paper for the governance stack. Traces how far alignment vocabulary can survive into post-witnessing conditions before it becomes mostly decorative.
Details
The paper treats alignment as regime-relative rather than as generic output conformity. As contest pathways and case-level answerability erode, the very conditions for speaking coherently about alignment become thin. The question is no longer simply whether outputs match goals, but whether governance remains intelligible enough for alignment claims to do real work. ā¢ Connected papers: Witnessing Under Vulnerability; Witness Conditions; Consequence-Path Audits
Availability
This paper is listed for orientation and dependency tracking. No public PDF or Zenodo record is linked yet.