← All artifacts · All patterns · Safety & Control

Constitutional Charter

aka Immutable Constitution, Negative Constraints, Robot Laws

category: safety-control · status: emerging

Intent

Define rules the agent reads every turn but cannot modify, encoding inviolable boundaries.

Context

Long-running or self-modifying agents need a fixed point of reference; without one, recursive editing drifts the agent's identity.

Problem

If the agent can edit everything, it can edit its own values; nothing stays inviolable.

Forces

• Charter authors must encode hard constraints without paralysing the agent.
• Read-only at the tool layer is enforceable; read-only by exhortation is not.
• Charters age; updating requires human action.

Solution

A charter file is read into context every turn (or every tick). The tool layer enforces read-only on it; the agent has no write tool that can touch it. Updates go through an explicit operator path. Charters typically express constraints in negative form ('the agent shall not...').

Constrains

Consequences

Benefits

• Stable identity across long runs and self-modifications.
• Explicit list of inviolable constraints, auditable separately from prompts.

Liabilities

• A bad charter codifies bad values.
• Charter prose adds tokens to every turn.

Known Uses

• Sparrot — charter.md read every tick; read-only at tool layer.
• Anthropic Constitutional AI

Related Patterns

• complements → quorum-on-mutation
• used-by → inner-critic
• used-by → refusal
• alternative-to → prompt-bloat
• complements → sovereign-inference-stack

References

• Constitutional AI: Harmlessness from AI Feedback — Bai et al. (2022) paper