Entrenchment and the Corrigibility Paradox
The 1933 Enabling Act shows a correction channel used, with complete formal validity, to abolish itself; postwar Germany's Article 79(3) (Ewigkeitsklausel) responds by placing the correction channel's own integrity conditions, not any current policy, outside the ordinary amendment process.
What decision changes?
When designing a corrigibility mechanism, entrench the meta-level — the integrity conditions of the correction machinery and the bearer floor — not the current value bundle, and only where an enforcement handle (a guardian with real power) exists to make the entrenchment more than ceremonial.
In March 1933, the German Reichstag voted — by the required supermajority, in a session conducted according to law — to transfer its own legislative power to Hitler’s cabinet. Every procedural requirement of the Enabling Act was satisfied. This is the sharpest case in the historical record of a correction channel being used, with complete formal validity, to abolish itself. No audit that checked only procedure could have caught it, because the procedure was precisely the weapon. The lesson generalizes uncomfortably: a system’s capacity to be changed includes, by default, the capacity to be changed into something that can never be changed again.
Postwar Germany’s constitutional answer is the institutional resolution of what AI researchers now call the corrigibility paradox — a fully correctable system can be corrected into permanent non-correctability, so some part of the system must sit outside the channel that updates everything else. Article 79(3) of the Basic Law, the Ewigkeitsklausel or “eternity clause,” makes human dignity and the democratic, federal, rule-of-law core of the constitution formally unamendable by any procedure — including the ordinary supermajority amendment process that the Enabling Act had exploited.
Two features of this design deserve careful attention. First, notice what it entrenches. Not any current policy, tax rate, or moral consensus — those must remain revisable, because freezing them would be its own catastrophe. What is placed beyond amendment is the meta-level: the integrity of the correction machinery itself and the floor of who counts. And entrenchment is not a free win. It is itself an anti-correction device, permanently narrowing what the ordinary channel may touch; it buys survival of the protected coordinate at a real price, which is why choosing the coordinate carefully matters so much. The AI translation is direct: what belongs outside a system’s own self-modification reach is the correction channel’s integrity conditions — not the current value settings, which must stay updatable.
Second, an eternity clause without an enforcement handle behind it is ceremonial. Article 79(3) has held because Germany’s Federal Constitutional Court can act on it and because the postwar order includes instruments (such as the power to ban anti-constitutional parties) that give it teeth. Formally similar clauses in jurisdictions with weaker courts have not reliably held. Which raises the obvious regress — who guards the guardian? — a question the dual-mandate-genesis case takes up directly.
One of eleven historical case studies in Institutional Genesis, Memory, and Decay — see the overview for the full life-cycle map, or read the complete appendix.
What would count as evidence?
Every procedural requirement of the 1933 Enabling Act was satisfied — the valid-reference-process condition failed completely while every surface procedural coordinate passed, which no audit that only checks procedure could have caught; Article 79(3) has held because the Federal Constitutional Court and militant-democracy instruments can act on it, unlike formally similar clauses in weaker judiciaries.