Hugging Face's logo

Datasets:
csoai
/
evidence-index

Languages:
English
Size:
n<1K
Tags:
provenance
evidence
ed25519
opentimestamps
measurement
License:
Dataset card Files
xet
 Community
Dataset Viewer
Duplicate
The dataset viewer is not available for this split.
Cannot load the dataset split (in streaming mode) to extract the first rows.
Error code:   StreamingRowsError
Exception:    CastError
Message:      Couldn't cast
files: list<item: struct<git_oid: string, lfs_sha256: null, path: string, size: int64>>
  child 0, item: struct<git_oid: string, lfs_sha256: null, path: string, size: int64>
      child 0, git_oid: string
      child 1, lfs_sha256: null
      child 2, path: string
      child 3, size: int64
repo: string
revision: string
schema: string
signature_checks: null
item: string
byte_checked_paths: list<item: string>
  child 0, item: string
to
{'byte_checked_paths': List(Value('string')), 'item': Value('string'), 'revision': Value('string'), 'schema': Value('string'), 'signature_checks': Json(decode=True)}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/utils.py", line 147, in get_rows_or_raise
                  return get_rows(
                      dataset=dataset,
                  ...<4 lines>...
                      column_names=column_names,
                  )
                File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
                  return func(*args, **kwargs)
                File "/src/services/worker/src/worker/utils.py", line 127, in get_rows
                  rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
                File "/src/services/worker/src/worker/utils.py", line 483, in safe_iter
                  yield from ds.decode(False) if ds.features else ds
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2840, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2373, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2398, in _iter_arrow
                  for key, pa_table in self.ex_iterable._iter_arrow():
                                       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 536, in _iter_arrow
                  for key, pa_table in iterator:
                                       ^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 419, in _iter_arrow
                  for key, pa_table in self.generate_tables_fn(**gen_kwags):
                                       ~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 343, in _generate_tables
                  self._cast_table(pa_table, json_field_paths=json_field_paths),
                  ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
                  pa_table = table_cast(pa_table, self.info.features.arrow_schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2378, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2306, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              files: list<item: struct<git_oid: string, lfs_sha256: null, path: string, size: int64>>
                child 0, item: struct<git_oid: string, lfs_sha256: null, path: string, size: int64>
                    child 0, git_oid: string
                    child 1, lfs_sha256: null
                    child 2, path: string
                    child 3, size: int64
              repo: string
              revision: string
              schema: string
              signature_checks: null
              item: string
              byte_checked_paths: list<item: string>
                child 0, item: string
              to
              {'byte_checked_paths': List(Value('string')), 'item': Value('string'), 'revision': Value('string'), 'schema': Value('string'), 'signature_checks': Json(decode=True)}
              because column names don't match

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

CSOAI evidence index

One signed, machine-readable list of every piece of evidence CSOAI (Council of AI, councilof.ai) has published on the sources below, so a person or an agent can discover it and verify it in one place. Built 2026-09-25T14:13:10Z; 212 items.

This is an index, not a measurement and not a certification. It proves which bytes were published and which of them carry a signature that verifies; it does not make any indexed claim true, and inclusion endorses nothing.

Files

file what
index.json the index (schema csoai.evidence-index/0.1), sha256 951a813866ffc745f6a1f0bb9210865099f1aa5c232cb520d4439360b9e9d2e9
index.signed.json Ed25519 signature over a payload pinning index.json by sha256, did:web:csoai.org#board-attestation-1, signed 2026-09-25T14:13:38.191Z
index.json.ots / index.ots.json OpenTimestamps proof over sha256(index.json): PENDING_CALENDAR_COMMITMENT (https://alice.btc.calendar.opentimestamps.org, https://bob.btc.calendar.opentimestamps.org, https://finney.calendar.eternitywall.com)
index.md the same index for humans
verify.py a stranger's one-command verifier (Apache-2.0)
build_evidence_index.py, render_evidence_index.py the code that built this (Apache-2.0)
manifests/ the per-repo file manifests whose sha256 is each Hugging Face item's content_sha256
checks/ per-repo lists of the files downloaded and re-hashed and the signature files checked; each bound to index.json by check.checks_sha256
enumeration/owners.json the PyPI / npm owner enumeration the package items were selected from
verify.log the output of python3 verify.py at build time

Verify 

pip install cryptography opentimestamps huggingface_hub
huggingface-cli download csoai/evidence-index --repo-type dataset --local-dir evidence-index
cd evidence-index && python3 verify.py

verify.py sends its own named User-Agent (csoai.org answers 403 to Python's default one). By hand:

import json, hashlib, base64, urllib.request
from cryptography.hazmat.primitives.asymmetric import ed25519
s = json.load(open("index.signed.json"))
c = json.dumps(s["payload"], sort_keys=True, separators=(",", ":"), ensure_ascii=False).encode()
assert hashlib.sha256(c).hexdigest() == s["signature"]["payload_sha256"]
assert hashlib.sha256(open("index.json", "rb").read()).hexdigest() == s["payload"]["artifact"]["sha256"]
did = json.load(urllib.request.urlopen(urllib.request.Request("https://csoai.org/.well-known/did.json", headers={"User-Agent": "csoai-dataset-verify/1.0 (+https://huggingface.co/csoai)"})))
x = [m for m in did["verificationMethod"] if m["id"].endswith("#board-attestation-1")][0]["publicKeyJwk"]["x"]
ed25519.Ed25519PublicKey.from_public_bytes(base64.urlsafe_b64decode(x + "==")).verify(bytes.fromhex(s["signature"]["sig_ed25519"]), c)

In words, the signature: in index.signed.json serialise payload as JSON with keys sorted recursively, no whitespace, UTF-8 literal. Its sha256 must equal signature.payload_sha256; payload.artifact.sha256 must equal sha256 of index.json; signature.sig_ed25519 (hex) must verify over those bytes under the #board-attestation-1 key (publicKeyJwk.x, base64url) in https://csoai.org/.well-known/did.json. Change one byte and it must fail.

The timestamp: ots upgrade index.json.ots && ots verify index.json.ots. At publication it was a pending calendar commitment — calendars promised Bitcoin inclusion; it is not a Bitcoin attestation until ots verify says so against the chain.

Without a Bitcoin node (a full ots verify needs one), check an upgraded proof offline and compare the merkle root with a public block explorer (two independent ones is better):

# (a pending proof must first be upgraded: ots upgrade <proof>)
ots --no-bitcoin verify -f index.json index.json.ots
# -> "To verify manually, check that Bitcoin block H has merkleroot M"
B=$(curl -s -A "csoai-dataset-verify/1.0 (+https://huggingface.co/csoai)" https://mempool.space/api/block-height/H)
curl -s -A "csoai-dataset-verify/1.0 (+https://huggingface.co/csoai)" https://mempool.space/api/block/$B | python3 -c 'import sys,json;print(json.load(sys.stdin)["merkle_root"])'
# must print M; repeat against https://blockstream.info/api/...

verify.py at build time:

INDEX index.json sha256=951a813866ffc745f6a1f0bb9210865099f1aa5c232cb520d4439360b9e9d2e9 signature=VERIFIED key=did:web:csoai.org#board-attestation-1 bound_to_index=True
RESULT pass=212 drifted=0 fail=0 not_checked=0 of 212 items

What it contains

kind items signature state OTS state
trust-anchor 1 NOT_APPLICABLE 1 none 1
board 2 VERIFIED 2 none 2
surface 3 UNSIGNED 2, VERIFIED 1 none 3
signed-record 32 VERIFIED 32 bitcoin-attested 24, pending 8
dataset 120 NOT_CHECKED 2, PARTIAL 6, UNSIGNED 104, VERIFIED 8 PARTIAL 2, bitcoin-attested 1, mixed 9, none 108
space 39 UNSIGNED 39 none 39
model 2 UNSIGNED 2 none 2
package 13 UNSIGNED 13 none 13

Counts of indexed objects, not measurements. Signed records verified under #board-attestation-1: 32 of 32.

What it does not show

  • That any item is correct, safe, or complete: an index proves which bytes were published and who signed them, not that their claims are true.
  • Any certification or endorsement: CSOAI measures; it does not certify, and inclusion here endorses nothing.
  • Evidence not reachable from the enumerated sources (see enumeration.not_enumerated). A source absent from this list leaves no trace here.
  • That live surfaces still serve these bytes: they are fetched LIVE and pinned only to this build's fetch.
  • Bitcoin-anchored time for any proof whose state is pending.

Not enumerated in this version:

  • GitHub: the estate's GitHub accounts are flagged and the repos are not publicly readable; nothing from GitHub is indexed.
  • Zenodo DOIs (e.g. 10.5281/zenodo.21991104 cited by /api/gspc): not enumerated in this version.
  • Kaggle, MCP registries, Smithery, other directories: listings, not evidence CSOAI publishes; not indexed.
  • Files inside the councilof.ai site beyond the surfaces listed above.
  • Signature files inside Hugging Face repos whose names do not match (signed|attest|provenance_signed|.sig$).

Package inclusion rule (rule v2 (2026-09-25; supersedes v1, which included every name candidate))

CSOAI (the measurement body) and MEOK (training and gamification) are separate estate entities. A package is indexed only when the registry bytes name CSOAI as its publisher, never because its name contains 'csoai'.

  1. Candidate: a PyPI project owned by, or an npm package maintained by, the estate registry account, whose name contains 'csoai' or 'gspc' (plus the named npm seed). The name only proposes; it never includes.
  2. Registry ownership (necessary, not sufficient): the PyPI Owner role (XML-RPC package_roles) or the npm maintainers list must include the estate account. These accounts are shared by CSOAI, MEOK and CSGA Global, so this step cannot establish CSOAI as publisher.
  3. npm scope: a scoped package is excluded unless its scope is a CSOAI scope (there are none). @meok-labs is MEOK's npm org; @csgaglobal is CSGA Global's.
  4. Declared publisher of the pinned release (PyPI author, author_email, maintainer, maintainer_email, home page; npm author): any MEOK or CSGA marker excludes it as another entity's package, including co-branded strings such as 'MEOK AI Labs (CSOAI LTD)'.
  5. Positive CSOAI identity required: the declared publisher names CSOAI Ltd / csoai.org / Council of AI / councilof.ai, or the declared source repository is under github.com/CSOAI-ORG. Otherwise the package is excluded as PUBLISHER_UNDECLARED (not evidence it is not CSOAI's; the registry bytes do not say).

Estate registry accounts: npm csga_global, pypi nicholastempleman. Decisions: EXCLUDED_OTHER_ENTITY 4, EXCLUDED_PUBLISHER_UNDECLARED 2, INCLUDED 13.

registry package version decision reason declared publisher
pypi crewai-csoai 0.1.0 INCLUDED declared publisher names CSOAI author_email: CSOAI Ltd *@csoai.org
pypi csoai 0.2.2 INCLUDED declared publisher names CSOAI author: Council of AI (CSOAI LTD, UK 16939677)
pypi csoai-affective-safety 0.1.0 EXCLUDED_PUBLISHER_UNDECLARED no publisher and no CSOAI-ORG source declared in the release metadata (none declared)
pypi csoai-claimguard 0.1.0 INCLUDED declared publisher names CSOAI author_email: CSOAI Ltd *@csoai.org
pypi csoai-council-ledger 0.1.0 INCLUDED declared publisher names CSOAI author_email: CSOAI Ltd *@csoai.org
pypi csoai-defoneos-isr-mcp 1.0.0 EXCLUDED_OTHER_ENTITY declared publisher names MEOK author_email: "MEOK AI Labs (CSOAI LTD)" *@meok.ai
pypi csoai-defoneos-mcp 1.0.6 INCLUDED declared publisher names CSOAI author_email: CSOAI Ltd *@csoai.org
pypi csoai-governance-crosswalk-mcp 1.0.18 EXCLUDED_OTHER_ENTITY declared publisher names MEOK author_email: MEOK AI Labs *@meok.ai
pypi csoai-gspc 0.2.20260915 INCLUDED declared publisher names CSOAI author_email: CSOAI Ltd *@csoai.org
pypi csoai-scitt-ts 0.1.0 INCLUDED declared publisher names CSOAI author: CSOAI LTD
pypi inspect-gspc-scitt 0.1.0 EXCLUDED_PUBLISHER_UNDECLARED no publisher and no CSOAI-ORG source declared in the release metadata (none declared)
pypi inspect-gspc-scorer 0.1.0 INCLUDED declared publisher names CSOAI author: CSOAI LTD
pypi langchain-csoai 0.1.0 INCLUDED declared publisher names CSOAI author_email: CSOAI Ltd *@csoai.org
pypi llama-index-tools-csoai 0.1.0 INCLUDED declared publisher names CSOAI author_email: CSOAI Ltd *@csoai.org
npm @csgaglobal/csoai-governance 1.0.0 EXCLUDED_OTHER_ENTITY npm scope @csgaglobal is not a CSOAI scope (CSGA Global) author: CSGA Global — Cyber Security Global Alliance
npm @meok-labs/csoai 0.1.0 EXCLUDED_OTHER_ENTITY npm scope @meok-labs is not a CSOAI scope (MEOK) (none declared)
npm csoai-gspc-mcp 0.2.2 INCLUDED declared source repository is under github.com/CSOAI-ORG (release published by github trusted publisher) (none declared)
npm csoai-lib2b 0.1.0 INCLUDED declared publisher names CSOAI author: CSOAI, *@csoai.org
npm csoai-x402-rail-client 0.1.0 INCLUDED declared publisher names CSOAI author: Nicholas Templeman, *@csoai.org

Every item carries its own does_not_show list, its licence, its published date, and the exact check behind its signature and timestamp state.

Contact, objections and re-checks

  • Contact: nicholas@csoai.org (Council of AI / CSOAI). Please name this dataset and the repository, package or item you mean.
  • Ask for a re-check. Reply with the repository, package or item. We re-measure it with the same read-only instrument and run a tamper control beside it (a deliberately altered copy must fail the same check, or the re-check does not count). A dated re-check record is published only with your knowledge. Published rows are never edited in place: a correction is a new, signed file that supersedes the old one, and both stay visible.
  • Object, or opt out of future reads. Name the repository, package or item. It is added to the exclusion list (scripts/census/probe-exclusions.json in the census code, dated, public), and the next run honours it before sending any request and records the skip by name. The reader behind this dataset (build_evidence_index.py) does not yet read that list in code; until it does, an entry is honoured by hand before the next run, and the skip is still recorded by name.
  • Objections are recorded. Every objection, opt-out and re-check request is logged with its date and what was done; nothing is removed silently.

Citation 

@misc{csoai_evidence_index_2026,
  author       = {{Council of AI (CSOAI)}},
  title        = {CSOAI evidence index},
  year         = {2026},
  publisher    = {Hugging Face},
  howpublished = {\url{https://huggingface.co/datasets/csoai/evidence-index}}
}

Licence

Index data CC-BY-4.0 (attribute: Council of AI, CSOAI Ltd 16939677, councilof.ai). verify.py and the build scripts Apache-2.0. Each indexed item keeps its own licence, stated per item; this index relicenses nothing.

Measurement-index chain

measurement-index/<date>/ holds CSOAI's daily measurement index as a hash chain: every day's index commits to the previous day's (SHA-256 and root) and declares any missed day. Each index is signed (did:web:csoai.org#board-attestation-1) and OpenTimestamps-stamped, and the chain starts at a signed genesis record (26 Sep 2026). How to verify it: measurement-index/README.md.

How to cite

CSOAI Ltd (Council of AI). CSOAI evidence index. 2026. Hugging Face dataset csoai/evidence-index. https://huggingface.co/datasets/csoai/evidence-index

@misc{csoai_evidence_index,
  title        = {CSOAI evidence index},
  author       = {{CSOAI Ltd}},
  year         = {2026},
  howpublished = {Hugging Face dataset, https://huggingface.co/datasets/csoai/evidence-index},
  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
9,543
Total file size:
8.33 MB

Space using csoai/evidence-index 1