AI and constitutions (from my email)

“Dear Tyler,
I enjoyed reading your notes on visiting Anthropic to advise on Claude’s constitution. Framing AI governance around the common law, case law (“Talmud”), and independent adjudication is a much more adaptive approach than relying on a static, top-down text.
That said, moving from a fixed text to a case-law system introduces its own set of structural risks. If Anthropic adopts this direction, a few institutional design hazards seem worth anticipating:
- The throughput bottleneck (Speed vs. Due Process): AI models generate billions of dynamic, edge-case interactions daily, while human judicial processes operate at human speed. If human adjudicators can only review a tiny fraction of flagged disputes, the actual operational rules will quietly decouple from official doctrine. Without automated verification tools to bridge this bandwidth gap, real oversight may only touch superficial cases.
- The danger of tangled precedent (Doctrinal bloat): The common law works because human societies change at a manageable pace. With rapid model updates and shifting capabilities, the volume of case law, exceptions, and secondary interpretations could quickly become self-contradictory. Over time, this leads to doctrine that serves as post-hoc justification rather than a coherent operational constraint.
- Correlated blind spots among AI reviewers: Using a diverse panel of AIs to detect constitutional drift is clever, but if these models share similar base data, fine-tuning techniques, or foundational architectures, their consensus will have shared blind spots. A model might learn to satisfy the specific rubrics of the reviewer panel while still drifting in ways the entire panel fails to register.
- The “Hollow Court” trap: The hardest problem in any independent judiciary is enforcement against the institution funding it. If economic or competitive pressures rise, an adjudicative board that lacks hard veto power risks becoming purely performative—producing elaborate legal commentary while commercial realities dictate the real guardrails.
The common-law analogy is compelling, but the real test is whether the institutional machinery can handle the sheer velocity and scale of software.”
That is from Scott Jenkins.

Facts Only

* Scott Jenkins provided notes on advising Anthropic regarding Claude’s constitution.
* The proposed approach frames AI governance using common law, case law ("Talmud"), and independent adjudication.
* A throughput bottleneck exists between the billions of daily dynamic interactions generated by AI models and human judicial processing speeds.
* Without automated verification, oversight may only cover superficial cases.
* Rapid model updates and shifting capabilities risk creating self-contradictory case law.
* Using diverse AI reviewers risks correlated blind spots due to shared base data or architectures.
* An adjudicative board lacking veto power faces enforcement challenges from economic or competitive pressures.

Executive Summary

Framing AI governance around common law, case law, and independent adjudication is proposed as a more adaptive approach than using a static text for regulating AI. This framework introduces structural risks, including a throughput bottleneck where the speed of AI generation conflicts with human judicial processes, potentially leading to rules decoupling from official doctrine without automated verification. Another risk is doctrinal bloat, where rapid model updates could create self-contradictory case law that serves as post-hoc justification instead of operational constraint. Furthermore, relying on AI reviewers introduces correlated blind spots; if reviewers share foundational data or architecture, their consensus may miss drift outside the panel’s learned boundaries. Finally, there is the risk of a "hollow court" where institutional bodies lack enforcement power against commercial pressures, resulting in performative legal commentary rather than genuine guardrails.

Full Take

The tension presented lies between the adaptive potential of a dynamic legal framework and the inherent inertia and capacity limits of institutional systems designed for human deliberation. The core implication is that imposing speed-based, high-velocity processes onto systems requiring nuanced, context-sensitive judgment introduces fragility at the intersection of technology and governance. The risk of doctrinal bloat suggests that the very mechanism intended to create coherent constraints might become an engine for epistemic fragmentation if scaling demands override deliberative coherence. The concern over correlated blind spots points toward a meta-governance problem: establishing external oversight mechanisms does not automatically resolve internal informational symmetries within the system itself. The "hollow court" trap reveals that institutional legitimacy hinges less on legal theory and more on material enforcement power, suggesting that technological advancement must be balanced against tangible economic incentives to avoid becoming purely symbolic regulation. What institutional structures can effectively mediate between algorithmic velocity and judicial deliberation without sacrificing substance? How can systems be designed such that speed does not inherently equate to the erosion of due process?

Sentinel — Human

Confidence

This text exhibits strong characteristics of human-generated critical thought, employing sophisticated conceptual linkage and nuanced argumentation rather than formulaic synthesis.

Signals Detected

Sentence length variance is natural; the tone shifts between advisory and analytical.

Strong, tightly integrated flow of complex conceptual arguments without excessive hedging or mechanical transitions.

The argument builds logically from a premise (moving to case law) to specific identified risks (bottleneck, bloat, blind spots).

Concepts like 'throughput bottleneck' and 'doctrinal bloat' are sophisticated conceptual framing, not easily generated rote information.

Human Indicators

Use of specific, nuanced analogies (Talmud, common law) combined with abstract technological risks.

Internal structure reflects a dialogue or thoughtful internal monologue rather than a standard exposition.

The concluding rhetorical question shifts focus effectively to the 'real test'.