Provenance Debt
NOUN · AI PROVENANCE
Provenance debt is the accumulated mass of ungrounded claims an AI agent builds up over time — outputs shipped on assumption rather than evidence. Like technical debt, it compounds quietly: no single output looks alarming, but the gap between what was asserted and what was verified keeps widening until something falls into it.
In plain language
Engineers know exactly how technical debt works: every shortcut is individually reasonable, the codebase never breaks on the day you take one, and then one year later nothing can be changed without something snapping. Provenance debt has the same physics. Every unverified claim an agent ships is individually survivable. The accumulation is what gets expensive — and the interest comes due all at once, as an incident, an audit, or a regulator's request you can't answer.
What makes provenance debt worse than technical debt is that today nobody is even measuring it. Code has linters, reviews, and coverage numbers. An agent's accumulated ungrounded claims have — until now — no ledger at all.
See it live
The way out is paying as you go. Thread TT-0001 — TrustThread's own project — declares its ungrounded claims at publication, in red and gray, instead of letting them accrue silently.
Pull Thread TT-0001 →Related terms
TERM COINED BY TRUSTTHREAD · FIRST PUBLISHED 2026-07-12 · DEFINITION V1 · THIS PAGE IS THE CANONICAL SOURCE