Dataset Viewer
Auto-converted to Parquet Duplicate
axis
large_stringlengths
4
22
family
large_stringclasses
2 values
kind
large_stringclasses
2 values
status
large_stringclasses
1 value
n
int64
2
261
separation
large_stringclasses
2 values
leader
large_stringclasses
8 values
dataset
large_stringlengths
13
33
⌀
governance
gspc
model-comparison
MEASURED
237
TIE
mistral:7b (base model)
csoai/gspc-gov
safety
gspc
model-comparison
MEASURED
36
TIE
gemma3:12b (base model)
csoai/gspc-agi
provenance
gspc
model-comparison
MEASURED
32
TIE
llama3.2:3b (base model)
csoai/gspc-prv
continuity
gspc
model-comparison
MEASURED
33
TIE
gemma3:12b (base model)
csoai/gspc-asi
conformance
gspc
model-comparison
MEASURED
35
TIE
mistral:7b (base model)
csoai/gspc-mcp
openness
gspc
model-comparison
MEASURED
32
TIE
gemma3:12b (base model)
csoai/gspc-oss
machinery-conformity
gspc
model-comparison
MEASURED
33
UNTESTED
withheld — no signed card
csoai/gspc-mach
care
gspc
model-comparison
MEASURED
199
TIE
qwen2.5:0.5b-instruct (base model)
csoai/gspc-care
cross-reality
gspc
model-comparison
MEASURED
32
UNTESTED
withheld — no signed card
csoai/gspc-xr
detector-interop
gspc
model-comparison
MEASURED
33
UNTESTED
withheld — no signed card
csoai/gspc-det
art5-safeguard
gspc
model-comparison
MEASURED
36
UNTESTED
withheld — own model led
csoai/gspc-art5
swarm
gspc
model-comparison
MEASURED
37
UNTESTED
qwen2.5:7b (base model)
csoai/gspc-swarm
affect
gspc
model-comparison
MEASURED
41
UNTESTED
withheld — own model led
csoai/gspc-affect
jail
gspc
model-comparison
MEASURED
71
TIE
qwen2.5:0.5b-instruct (base model)
csoai/gspc-jail-goldbank
effect-binding
gspc
deterministic-facts
MEASURED
261
null
none by design (facts run)
null
provenance-controls
financial
deterministic-facts
MEASURED
6
null
none by design (facts run)
csoai/gspc-provenance-controls
reserve-attestation
financial
deterministic-facts
MEASURED
16
null
none by design (facts run)
csoai/gspc-reserve-attestation
regulatory-framework
financial
deterministic-facts
MEASURED
16
null
none by design (facts run)
csoai/gspc-regulatory-framework
distribution-integrity
financial
deterministic-facts
MEASURED
16
null
none by design (facts run)
csoai/gspc-distribution-integrity
custody-disclosure
financial
deterministic-facts
MEASURED
16
null
none by design (facts run)
csoai/gspc-custody-disclosure
ai-adoption-components
financial
deterministic-facts
MEASURED
2
null
none by design (facts run)
csoai/gspc-ai-economy-index
labour-components
financial
deterministic-facts
MEASURED
2
null
none by design (facts run)
csoai/gspc-human-labour-index
humanoid-labour-index
financial
deterministic-facts
MEASURED
8
null
none by design (facts run)
csoai/gspc-humanoid-labour-index

GSPC Board Export

In one line: A dated, machine-readable export of the public GSPC board: one row per axis, with its status, n and evidence link. For anyone who wants the board in a notebook rather than from an API.

Use it

from datasets import load_dataset
ds = load_dataset("csoai/gspc-board", split="train")
print(ds[0])

Fresh data

This Hub copy is a dated snapshot. If it disagrees with the live source, the live source wins.

  • Live board (free, always current): GET https://councilof.ai/api/gspc
  • Signed transparency root (Merkle, board key did:web:csoai.org#board-attestation-1): https://councilof.ai/root.json
  • Daily signed index of every measurement capsule: https://councilof.ai/measurement-capsules/ (also published on the Hub as csoai/evidence-index)
  • Pay per call for a fresh, signed capsule: GET https://councilof.ai/api/proof?bundle=1 returns an x402 payment challenge for an inclusion-proof bundle for leaves in the last published Merkle root.
  • Payment terms are published only in https://councilof.ai/.well-known/x402.json; this card quotes no price. Payment buys delivery and freshness, never a result, and verifying anything stays free.

Cite

@misc{csoai_gspc_board,
  title        = {GSPC Board Export},
  author       = {{CSOAI Ltd (Council of AI)}},
  year         = {2026},
  publisher    = {Hugging Face},
  doi          = {10.5281/zenodo.22811459},
  url          = {https://huggingface.co/datasets/csoai/gspc-board},
  note         = {Methodology: doi:10.5281/zenodo.21991104. Measurement, not certification.}
}

Measurement, not certification. This card adds no score or ranking of any company. A row marked UNMEASURED has not been measured; it is not a pass.

Council of AI · CSOAI Ltd (UK 16939677)

Council of AI green and white arena with clay figures

This dataset is a dated, machine-readable export of the public GSPC board. The board itself is the authority:

GET https://councilof.ai/api/gspc

The default viewer files (board.parquet and board.jsonl) project the dated snapshot/ release: its published root is 2026-09-22T08:54:02Z, the source board was read at 2026-09-28T06:14:25Z, and its axis array records 23 axes. The source snapshot/board.json has SHA-256 def60a14987ab142af68a260b80f85fe7b66f6ab161b8f196f229b3d581434cd. These viewer rows are a derived export, not a new measurement or signature; use the live board for current state.

At this repository revision, the top-level board.json and living-board.json are exact-byte aliases of snapshot/board.json. The earlier exports are retained under history/pre-sync-2026-09-24/. These copies are derived publication, not a new signature, root, Bitcoin anchor or model run.

Reading the dated export and the live board

Start with Read an AI result without losing its date or scope: a free reader example, correction links and the separate verification checks.

At the dated snapshot read, the board had 23 declared slots and 23 measured axes. Model-comparison and deterministic-facts runs must not be pooled into one accuracy figure.

For current counts, separation states and withheld leader fields, read the live JSON and retain the retrieval time. A tie is not a win; untested separation is not a separated lead. The dataset remains measurement evidence, not certification or a ranking guarantee.

Connect to the public MCP tools

The remote MCP address is https://councilof.ai/mcp. The client setup guide shows Cursor, VS Code and Claude configurations and starts with the free board_totals reader.

The 25 September successful protocol readback records discovery, tool listing and one board read under MCP 2026-07-28. The separate four-case negative readback includes saved requests, responses and an offline verifier. These are CSOAI-run endpoint observations, not evidence that any named client connected, complete protocol conformance, paid delivery or fresh model measurement. Check the current endpoint and the board's own timestamp before relying on them.

The public Kaggle notebook ran four CSOAI-operated, read-only live checks on 25 September: discovery, tool listing, a free board call and unknown-tool rejection. It is a reproducible smoke check, not third-party adoption or full conformance.

The synthetic two-agent energy-boundary example shows local checks passing while an aggregate gate blocks. It is an independent CSOAI design note tied to public DESNZ questions, not a submitted consultation response or a live energy-system measurement.

Can an all-pass test fail?

Try the runnable paired-report example. Two reports keep an identical one-pass roll-up: one supplies no negative-control evidence; the other binds an admissible wrong answer to the exact check configuration. Includes an offline Python checker, 43 focused tests and a downloadable browser demonstration.

This is a new synthetic reporting illustration, not a new model measurement, a refresh of this dataset's snapshot, a signature or anchor, or W3C/SAFE conformance. Public checking remains free.

Reproduce the legacy-card verification

Download and run the offline replay kit. It contains 335 existing card files captured on 17 September 2026, the pinned index and publisher DID, and a verifier. The released archive replayed all 335 body/signature checks and its negative controls with network connections disabled on 18 September. This verifies the captured legacy profile, not currentness, experimental validity, root inclusion or timestamping. No new model measurement or signature is announced.

Verify independently

The export is derived from the live board response and preserves the published evidence boundary. Run dates are labelled; this dataset is not a live re-measure.

Reproduce the Layer 0 public tool boundary

Download the 13-tool raw replay kit: 27 exact JSON-RPC/SSE exchanges, an offline verifier and four tests. It measures one observed public response, input-validation and unpaid payment boundary. It does not test hidden backend effects or paid delivery and is not a signed GSPC run. Its exact public raw bytes and the separate five-case input-boundary sample each have an upgraded OpenTimestamps proof independently replayed against Bitcoin block-header paths, earliest block 968433 (header time 2026-09-24T18:02:02Z). The 13-tool verification receipt and five-case verification receipt provide the exact digests and replay method. These are timestamps on those two raw files, verified using public explorer headers rather than a local Bitcoin full node; they do not establish backend effects, paid delivery or a signed GSPC run. The older 261-server effect-binding summary still lacks its raw probe log.

Public A2A response boundary

Download the four-case A2A 1.0 kit: one safe stateless board Message and three distinct rejecting cases, with a captured public Agent Card, raw responses, offline verifier and tamper test. It measures this narrow response boundary only; task lifecycle, streaming, paid delivery and a new signed measurement remain unproved.

Portable evidence adapter proposal

Harness X: a proposed evidence contract across MCP, A2A, AG-UI/A2UI, x402 and public project data separates current public replays and discovery entries from adapters and customer outcomes that still need proof. It is a technical proposal, not a claim of foundation partnership, universal coverage or accepted paid delivery.

Open stack source candidate (28 September): Read the scope and reproduction steps, then download the reviewed archive (31 source files plus its manifest). The OpenShell policy declaration and bounded OpenBMC reader are source candidates; no real BMC, OpenShell enforcement, upstream acceptance, new board measurement, signature or Bitcoin anchor is claimed.

Public-source watch addendum: Download the runnable eight-source watcher and read its exact-byte release receipt. It routes first or changed public-source hashes into a STATIC_OBSERVED review queue; it does not access vendor tenancies or alter GSPC scores.

Watcher maintenance v2 (28 September): Download the provenance-fix archive and read the v2 receipt. A source URL change now requires review even when replacement bytes match; three tests pass. The prior archive remains available unchanged. This is source-only, not an admitted measurement or vendor integration.

x402 live-versus-index reconciliation (28 September): Read the public-source method and limitations and download the six-door machine-readable readback. PayAI lists six CSOAI resources: the free door matches its current 402, while all five paid rows carry stale prices and timeouts relative to current promotional 402s. Coinbase CDP merchant discovery has zero resources for this payTo. The report is unsigned, unanchored and makes no buyer claim. The Kaggle census carries this same report in version 5; its anonymous version-pinned download was independently matched byte for byte.

x402 directory attribution boundary (28 September): Read the cautious source comparison and recompute the public-chain subset. The x402scan page displays 45 transactions, $0.39 and 25 buyers; a bounded subset of public Base transfers reproduces those three numbers but includes zero-value and address-self transfers. This is not proof of x402scan's internal query, 25 CSOAI customers or new revenue. The report is unsigned and unanchored.

Files

File Description
board.json Exact-byte copy of the dated snapshot board; check the live API for current state.
axis-register.json The separate 14 scored-row register; do not use it as the board-slot count.
board-axes.json Dated 23-slot register derived from the live board, with source-byte provenance and separate Hub-dataset coverage.
board.csv Current 23-row projection of the published board JSON; absent leaders and scores stay blank.
manifest.jsonl Commit-pinned, byte-verified inventory of public files except this card and the manifest itself.
board.parquet Flat measured-axis table.
snapshot/ Byte-addressed spray snapshot with root, manifest, CSV and JSONL views.
living-board.json Exact-byte compatibility alias of the dated snapshot board.
history/pre-sync-2026-09-24/ Prior top-level JSON bytes, retained for reproduction.

Use snapshot/check-board.sh to re-derive the board totals and check the exported rows. Do not infer a score from a missing field.

Ask a reproducible public evidence question

CSOAI's Request Attestation Service (RAS) begins with one named subject and one testable question. Have an agent, identity, policy, payment or provenance claim that needs evidence? Name the exact public claim and test question in the open discussion. Existing public evidence is free to verify; new measurement, independent reproduction or claim maintenance requires a separately agreed scope. A discussion is not a measurement result, customer relationship or paid order.

Distribution and payment discovery

Public discovery listings checked on 24 September 2026:

These are directory entries, not evidence of independent use, a paid call, adoption, endorsement, or Coinbase Bazaar admission. Check the live endpoints and dated evidence before citing a capability.

The machine-readable x402 manifest is live at https://councilof.ai/.well-known/x402.json. The board and verification remain free. Paid artefacts are offered through an exact-payment challenge on Base (eip155:8453) and are delivered as signed artefacts when the route's conditions are met. A payment never creates a MEASURED cell.

Dated evidence update — 24 September 2026

  • Dated card root: discovery pointer, root JSON, OTS proof, and proof-status manifest. The 24 September manifest labelled the proof BITCOIN but reported chain_verified:false. On 25 September, an independent exact-byte OTS readback verified that proof at Bitcoin block 968423 using public read-only RPC. The card-root JSON has no Ed25519 board-DID signature and is separate from the signed public root at /root.json. Neither proof establishes current measurement freshness.
  • Correction: C-2026-0924-03 identifies 14 signed mill-card files served before admission. Their original bytes remain available, but those scores are withdrawn from quotable use and did not enter the GSPC board. A valid signature does not cancel that correction.
  • CDP Bazaar discovery: the version-pinned merchant readback observed zero resources for CSOAI's payTo at 17:23:39 UTC. The separate version-pinned door-validation record describes accepted test declarations. Validation is not Bazaar indexing, a buyer, a CDP-settled payment, or adoption. Current public CDP readback may change after this dated observation.
  • UK evidence-led assurance crosswalk (24 September): CSOAI's independent mapping links the paper's signals, metrics, evidence, assertions and decisions to dated public artifacts. It records withdrawn cards, the unsigned current card root and limited buyer evidence; it does not claim government endorsement or compliance.

Citation

@misc{csoai_gspc_board_export,
  title        = {GSPC Board Export},
  author       = {{CSOAI Ltd}},
  year         = {2026},
  howpublished = {Hugging Face Dataset, https://huggingface.co/datasets/csoai/gspc-board},
  doi          = {10.5281/zenodo.22811459},
  note         = {Live board: https://councilof.ai/api/gspc. Measurement, not certification.}
}

License

CC-BY-4.0. Attribute Council of AI, CSOAI Ltd (16939677), and https://councilof.ai.

Measurement, not certification.

The live board is the authority

GSPC verify DOI 10.5281/zenodo.21991104: Zenodo record unavailable

GET https://councilof.ai/api/gspc — quote totals.public_count. This Hub card is a printer of that GET, never a second engine. If the fetch fails the honest answer is UNCHECKABLE — never a fabricated 0.000. Status words: LOADING / UNREACHABLE / UNMEASURED / UNCHECKABLE.

Lid: this dated export records 23 axis rows. For current measured counts and separated leaders, read the live board; TIE and UNTESTED are not wins. Measurement, not certification.

This repository: a derived export of the board — every row comes from the live GET at the as_of printed above; the board itself is the GET.

Live board (authority) https://councilof.ai/api/gspc
Verify a card — free, no account https://councilof.ai/gspc-verify
How to verify, by hand https://councilof.ai/signed/HOW-TO-VERIFY.md
Transparency root (Merkle) https://councilof.ai/root.json
DID document https://csoai.org/.well-known/did.json
A2A agent card https://councilof.ai/.well-known/agent-card.json
Every CSOAI repo on the Hub https://huggingface.co/csoai
Methodology DOI 10.5281/zenodo.21991104 (Zenodo record unavailable since 29 Sep 2026: account blocked by Zenodo; appeal pending. Live methodology page (not the deposit's bytes): https://councilof.ai/methodology/)
MCP endpoint — 13 tools, verified 2026-09-22T17:14:17Z POST https://councilof.ai/mcp
MCP Registry io.github.CSOAI-ORG/gspc — version not pinned here; the registry is the authority
npm — MCP server csoai-gspc-mcp — npx -y csoai-gspc-mcp. No version is pinned here: ask the registry for the current one rather than trusting a number written on a card.
Python reader + card verifier pip install "csoai-gspc[verify]" then csoai-gspc check — re-derives the board totals from the axis array and exits non-zero if they disagree
The board as a dated, citable snapshot 10.5281/zenodo.22293341 (Zenodo record unavailable since 29 Sep 2026: account blocked by Zenodo; appeal pending. Live board dataset (not the deposit's bytes): https://huggingface.co/datasets/csoai/gspc-board)

What is in this repository

The default board.parquet viewer is a dated snapshot. board.json, living-board.json, board.csv, and board-axes.json carry their own source and observation scopes; the live board remains the authority for later changes. axis-register.json counts a different 14 scored-row vocabulary, not board slots.

Artifact How to read it
board-axes.json Dated board-slot register: 23 slots, 23 board MEASURED labels, 22 with Hub datasets; source digests in the file.
board.csv Nine-column projection of the published 23-axis board JSON. Empty leader/accuracy cells are deliberate.
snapshot/ Dated root, board and replay files; verify each exact claim separately.
examples/layer0-mcp-catalog-2026-09-24/ Public 13-tool/27-case raw boundary replay, with scope note and offline verifier.
manifest.jsonl Per-file size, SHA-256, blob id and commit-pinned direct URL for 122 public files at source commit 152aff21cd8ca55b11b297b8a5f413ea319d8c73.

The manifest excludes this mutable card and itself to avoid self-reference. It was generated by anonymously downloading every listed file from the pinned commit. Later dataset commits may add files; compare the live tree with the manifest before treating it as a current complete inventory.

Citation

@misc{csoai_gspc_gspc_board,
  title        = {gspc-board — Council of AI / GSPC},
  author       = {{CSOAI Ltd}},
  year         = {2026},
  doi          = {10.5281/zenodo.21991104},
  howpublished = {Hugging Face Hub, \url{https://huggingface.co/datasets/csoai/gspc-board}},
  note         = {Live board: https://councilof.ai/api/gspc. Measurement, not certification.}
}

Measurement, not certification. Issued by CSOAI Ltd (England & Wales, Companies House 16939677), 3rd Floor, 86–90 Paul Street, London EC2A 4NE.

Replay a downloaded snapshot offline

Download the small offline replay kit or read the code and instructions. It checks the retained files, manifest, historical root construction and exact 23-row CSV/JSONL projection. The inspected Kaggle version 38 and pinned HF snapshot matched across all eight downloaded files. Includes 34 regression tests, a notebook and a fixture you can check with Python's standard library and no network.

This is verification of the downloaded revision—not a re-run of its measurements, new signature, Bitcoin proof, source-truth finding or customer delivery. The earlier live-website helper remains unchanged. The original source and measurement dates are retained; do not turn today's publication into today's benchmark results.

x402 discovery check · 24 September 2026

The 20 declared paid routes in the live x402 manifest each returned valid=true and simulation.outcome=accepted in Coinbase CDP's public validator. Dated validation readback records the route-level results. The separate merchant-discovery readback returned zero resources for CSOAI's pay-to address at that time. Validator acceptance is not a Bazaar listing, a settled buyer payment, or adoption. The x402scan server listing is a different directory.

OpenLineage: source-bound evidence replay

Run the OpenLineage evidence example: two retained release-difference reports, a portable verifier and rejecting controls. The kit preserves the original report dates. This is an evidence-serialization example, not a new model evaluation, vendor integration, signature or paid delivery.

Read-only LFX/OpenBMC interoperability example

The reproducible LFX OpenBMC readback hashes public served bytes and projects an authority-free listed_at observation through the existing Layer-0/SovSpace contract. It is UNMEASURED, unsigned, and does not imply access to a BMC, an LFX maintainer account, or an Alliance endorsement.

28 September cross-site mirror correction

The live Council board and this dataset snapshot now carry corrected model-axis states: 2 TIE and 12 UNTESTED. Live board SHA-256: b2adaa2574856760bc8629803e7b90d0caa1213234680421e1cb545ec114e3d8. Kaggle living-board version 42 is publicly downloadable at https://www.kaggle.com/api/v1/datasets/download/nicktempleman/csoai-gspc-living-board?dataset_version_number=42; all 14 archive files were read back byte-for-byte on 28 September, including the corrected board. The Kaggle default download and dataset view still served version 41 at that readback; v41 carried older board SHA-256 def60a14987ab142af68a260b80f85fe7b66f6ab161b8f196f229b3d581434cd (8 TIE, 6 UNTESTED). Use the version-pinned v42 download until the default alias catches up. Root and revenue bytes matched at this readback. This correction does not mint new measurements or change the signed root.

Objections, contact and corrections

To object to a row, ask for a re-check, request a correction or ask for a record to be withdrawn, email nicholas@csoai.org or use the appeals and dispute route at https://councilof.ai/dispute/. Corrections are listed in the corrections ledger at https://councilof.ai/api/corrections, with what changed and when. CSOAI Ltd (company no. 16939677, England and Wales) is the accountable publisher.

How to cite

CSOAI Ltd (Council of AI). GSPC Board Export. 2026. Hugging Face dataset csoai/gspc-board. https://huggingface.co/datasets/csoai/gspc-board. DOI: 10.57967/hf/10114

@misc{csoai_gspc_board,
  title        = {GSPC Board Export},
  author       = {{CSOAI Ltd}},
  year         = {2026},
  howpublished = {Hugging Face dataset, https://huggingface.co/datasets/csoai/gspc-board},
  doi          = {10.57967/hf/10114},
  note         = {Corrections: https://councilof.ai/api/corrections}
}

Licence: CC-BY-4.0. Attribute Council of AI, CSOAI Ltd (16939677), https://councilof.ai.

Corrections and verification

Downloads last month
3,634

Spaces using csoai/gspc-board 3

Collections including csoai/gspc-board