How Drift Gets Detected

A disclosure system faces a specific long-term risk that is different from failure. The risk is drift — the gradual transformation of a documentation tool into something that looks like certification, or enforcement, or ranking. Not through a single decision, but through accumulated small shifts over years.

Hedamo is designed to detect this.

The system maintains a set of drift indicators. If any of the following occur, the system has drifted and requires correction:

Disclosure scores are used as procurement eligibility criteria. A trade authority uses disclosure data as a market access gate. AI-generated outputs contain causal health language — "causes" or "improves" rather than "associated with." A government mandates disclosure rather than incentivizing it. A private entity gains control over schema evolution. Products are sorted or ranked against each other in any output.

Each of these represents a specific boundary. Individually, they are simple. Collectively, they form the perimeter that keeps a disclosure system from becoming a certification system.

The distinction matters because the institutional value of disclosure depends on it remaining disclosure. The moment a disclosure score becomes an eligibility threshold, it ceases to be a transparency tool and becomes a regulatory instrument — with all the liability, political exposure, and enforcement complexity that entails. The moment AI outputs make health claims, the system takes on clinical liability it was designed to avoid.

Drift detection is not a feature added after the system was built. It is part of the original architecture. The question "who keeps Hedamo from drifting into certification?" has a documented answer: published invariants, versioned schemas, auditable AI prompts, and explicit refusal hooks that operate at the architecture level, not the policy level.

The system watches itself. That is not a metaphor. It is a design requirement.