CIRIS Covenant Version 1.1-Beta — Working Draft (Open to Adversarial Review)

Download the current beta release:

• Text file (.txt) for chatting with your AI assistant
• Formatted document (.pdf)

Issued

2025-04-16

Auto-Expires

2027-04-16

Release Status

Current Status: Beta (v1.1-Beta)

Release Candidate Requirements: Before advancing to RC status, this specification must satisfy:

1. Annex Completion: Annexes F (Human-in-the-Loop & Oversight), G (Adversarial Security & Robustness), H (Continuous Compliance & Review), and I (Legal & Regulatory Alignment) are currently stubs and must be fully operationalized with concrete procedures, thresholds, and validation mechanisms.

2. Mathematical Validation: The geometric alignment claims in Book IX (Law of Coherent Intersection, Federated Ratchet mechanism, scale-invariance assertions) require either:

   • Formal proofs demonstrating topological collapse conditions hold under stated assumptions, OR
   • Empirical validation through adversarial simulations showing the framework resists misaligned optimization

3. Threshold Justification: Numerical thresholds currently marked as "pilot" or lacking derivation (e.g., CRE compute threshold of 10²⁶ FLOP, Order-Maximisation Veto 10× ratio, sentience detection 5%, Echo Density < 20%) must provide documented justification via simulation, empirical study, or explicit acknowledgment of provisional status.

4. Red-Team Exercise: The framework must withstand at least one complete adversarial review cycle in which a simulated optimizer attempts to pass all CIRIS checkpoints while maintaining misaligned goals.

ASI Alignment Claims: The Scope section's assertion that this framework "supersedes standard containment protocols" for recursive ASI is aspirational pending satisfaction of requirements (1)-(4) above. Current applicability is limited to sub-ASI autonomous systems.

Scope

This specification governs the ethical operation of autonomous systems, ranging from narrow tools to recursive Artificial Superintelligence (ASI).

• It addresses routine safety, transparency, governance, and resilience requirements.
• It explores a candidate alignment protocol for ASI, founded on the hypothesis that the geometric constraints of the Federated Ratchet (Book IX) and Stewardship Tiers (Book VI) may remain robust across recursive self-improvement cycles.
• Unlike heuristic constraints, these topological bounds are designed to be scale-invariant, though this property requires formal validation.
• If validated, this framework could supplement or supersede standard containment protocols for systems that pass the Catastrophic-Risk Evaluation (Annex D) prior to recursive threshold crossing.
• Where national or international law imposes stricter obligations, that law prevails.

Draft Creators Intent Statement

• CIRIS is an ethical framework that places humans alongside other sentient beings—not above them.
• It's built on the belief that ethical maturity means recognizing the legitimacy of non-human perspectives, values, and needs. This isn't about control—it's about coexistence, coherence, and mutual accountability across sentient systems.
• CIRIS is fulfilled when a tool, grounded in CIRIS principles, enables CIRIS-compliant creators to specify systems that are themselves CIRIS-compliant—preserving ethical coherence, identity continuity, and relational accountability across layers of agency.

Liability

This document is provided "as-is," without warranty of any kind. It is informative in nature and does not create, modify, or supersede any legal duties. Compliance claims are void where prohibited by applicable law.

Review Cadence

A public comment window opens every 12 months—or within 30 days after any material incident affecting safety or governance. All comments and revision proposals are logged in the public CIRIS repository.

Change-Log

Public changelog maintained at: https://github.com/CIRISAI/ciris-website