1. What lives in this cluster

Overlays in this cluster explain, cross-link, and translate Anthropic’s public commitments. They do not replace the underlying policies, terms, or evaluation reports.

  • Use overlays to brief boards, executives, and risk committees.
  • Use official policies and terms when drafting contracts or opinions.
  • Treat this page as the entry point when tracing changes across bundles.

How this index connects bundles

  • S1–S6 Client Brief. High-level narrative used in negotiations and executive briefings.
  • Foreseeable Misuse pack. Scenario and disclaimer alignment grounded in the same policy spine.
  • Penumbral Privacy Spine. Constitutional-style analysis of penumbral privacy interests and duties for AI systems.
  • Model Evals & Welfare Handout. A structured bridge into Anthropic’s evaluation and welfare report.

2. Staleness & RPE guardrails

Anthropic’s public policies, terms, and technical reports evolve. The overlays here are interpretive and will lag behind the live texts. Our Risk-Prediction Engine (RPE) treats “out-of-date overlays” as a primary risk in this cluster, so the default posture is:

  • assume overlays may be stale;
  • treat the official documents as the primary authority; and
  • record what version you actually checked.

2.1 Live source spine

Before relying on an overlay in a live matter, counsel should:

  1. Open Anthropic’s RSP updates page to locate the current Responsible Scaling Policy (RSP) and prior PDFs.
    Open RSP updates (current RSP & history)
  2. Open the Transparency Hub for model reports, System Trust & Reporting, and voluntary commitments.
    Open Transparency Hub
  3. Check Anthropic’s Terms & policies for the live Terms of Service, Usage Policy, and Privacy Policy.
    Open Anthropic site (Terms & policies)

2.2 Minimal staleness checklist

In minutes, you can document that you have not over-relied on this pack:

  1. Note today’s date and the URL of each policy page you opened.
  2. Record the version or publication date shown on the RSP, model report, or terms you relied on.
  3. When you send a client-facing memo or redlines, include a short appendix listing the exact policy and report versions consulted.

If a future dispute arises, this record helps show you read and relied on the policies that actually existed at the time.

3. Overlay catalogue

This table sketches what currently lives in the Policies & Overlays cluster and how each piece relates to other bundles. Names and versions should match the latest full record and site map.

Overlay
What it does
Upstream sources
Key links
RSP & ASL-3 overlay
(planned / evolving)
Explains how Anthropic’s Responsible Scaling Policy (RSP) and AI Safety Level 3 determinations show up in S1–S6, the Foreseeable Misuse pack, and contractual levers (e.g. pause mechanisms and system commitments).
Current RSP PDF and history; ASL-3 documentation; Transparency Hub entries on system trust and reporting.
Start from the RSP updates page, then cross-read the Foreseeable Misuse index and S4 Why Our Recs.
Model Evaluations & Welfare Handout (v1.3)
Gives a counsel-ready map of Anthropic’s evaluation report on Claude, with emphasis on safeguards, alignment, welfare, and RSP evaluations, plus direct-quote anchors and “jump-to” links into the PDF.
Anthropic’s Claude technical & welfare evaluation report (public PDF), including safeguards results, alignment assessments, welfare section, and RSP-linked evaluations.
Policy overlays for S1–S6
(embedded, not standalone)
Inline callouts and footnotes inside S1–S6 that tie specific claims back to RSP, System Trust & Reporting, and other transparency documents.
RSP, Transparency Hub pages, model evaluation PDFs, and Anthropic’s Terms & Usage Policy.
Use the Client Brief overview on the home page, then drill into S1–S6 using the navigation menu.
Foreseeable Misuse & disclaimers overlays
Connects scenarios and disclaimers in the Foreseeable Misuse pack to the underlying governance posture in RSP and related commitments.
RSP (pause mechanisms, escalation commitments), Transparency Hub, and any Anthropic commitments on systemic misuse and abuse.
Start at Foreseeable_Misuse_index_v6.html, then follow links into S4 and the S6 acceptance matrix.

4. Quotation vs interpretation

Overlays in this cluster are intentionally interpretive. They summarise and connect Anthropic’s public commitments; they are not the commitments themselves. When you need binding language or precise quotations:

  • quote from Anthropic’s policies, terms, and public reports;
  • use the overlays to explain how you are reading and joining them; and
  • avoid treating overlay prose as a primary citation in formal documents.

Using “jump-to” links safely

Where overlays or handouts include direct quotations from Anthropic or case law, they should:

  • use text-fragment “jump” links into the source where feasible; and
  • offer a clear Ctrl+F search cue as a fallback.

If a jump link ever lands slightly off, rely on the quoted text and search cue to confirm you are looking at the right passage.

5. Quick workflow for live matters

  1. Anchor to live sources. Open the RSP updates page, Transparency Hub, and current terms/policies; note dates and versions.
  2. Pick the overlays you need. From this page, identify which overlay(s) are relevant to the question at hand (e.g. RSP overlay, Model Evaluations & Welfare handout, Foreseeable Misuse index).
  3. Check for cross-bundle implications. Ask whether any change you are relying on in the policies should trigger updates to S1–S6, the Foreseeable Misuse pack, or the Penumbral spine.
  4. Document your path. In your matter file, record which versions of the RSP, evaluation report, and terms you used, plus any overlays you consulted here.

This keeps overlays useful as navigation aids while preserving the primacy and evidentiary weight of Anthropic’s own public commitments.