Resolves to the option matching the total pairwise structural similarity score of the endpoint snapshot (frameworks as published on 2027-06-30), per the pre-registered coding scheme published at Audit Motion 003|三大前沿安全框架,正在收斂還是分化?

Baseline (2026-07-18): 19/30.

Scoring: 5 dimensions (capability threshold definition, deployment-halt triggers, external verification role, disclosure obligations, amendment governance) × 3 framework pairs (Anthropic–OpenAI, Anthropic–GDM, OpenAI–GDM), each scored 0–2. Converge = ≥ 22 (baseline +3); Diverge = ≤ 16 (baseline −3); No significant change = 17–21.

The coding scheme, per-cell criteria, and baseline scores were published before market creation and are frozen. Endpoint scoring uses identical criteria. A second blind coder scores independently: if the two totals differ by ≤ 4, the primary coder’s score resolves the market (the blind score is published as a reliability record); if they differ by > 4, both scoresheets are published and the raw average resolves, with thresholds applied directly to the average (non-integer values between the thresholds resolve to No significant change).

Succession rule: if a framework is merged or renamed, its successor document (same legal entity, same function) is scored. If a framework goes non-public, it is frozen at its last public version and flagged in the audit report.

Baseline documents (archived 2026-07-18, SHA-256 in the pre-registration):

• Anthropic RSP v3.4: https://web.archive.org/web/20260718021024/https://cdn.sanity.io/files/4zrzovbb/website/0bacdc8440ea96e62a8766d99ebe1d4eea6d5f3a.pdf

• OpenAI Preparedness Framework v2: https://web.archive.org/web/20260718021059/https://cdn.openai.com/pdf/18a02b5d-6b67-4cec-ab64-68cdfbddebcd/preparedness-framework-v2.pdf

• Google DeepMind FSF v3.1: https://web.archive.org/web/20260718021230/https://storage.googleapis.com/deepmind-media/DeepMind.com/Blog/strengthening-our-frontier-safety-framework/frontier-safety-framework_3-1.pdf

This is Audit Motion 003 by Entropy Clinic (熵診所). Prior motions:

001: Audit Motion 001 — Anthropic at the AI R&D-4 Boundary (R1)

002: Audit Motion 002: When Governance Stops Being a Promise and Becomes a Metric

Audit Motion 002: When Governance Stops Being a Promise and Becomes a Metric

A pre-registered prediction on the emergence of recurring organizational governance metrics in frontier AI reporting.