{
  "schema": "csoai.memberships-updates/0.1",
  "as_of": "2026-09-22",
  "signed": false,
  "purpose": "The dated record of what changed in public/interop/memberships.json and why. One entry per change, newest first. Every entry names the manifest row it touches (or null when it touches the manifest as a whole) and carries the same kind of evidence the rows do: a public URL a stranger can open, or a dated mailbox record when the only evidence is private. This file exists so a reader can see that a row moved, when, and on what basis — a table that only ever shows its current state cannot be audited. scripts/memberships-check.mjs validates it and fails when an entry names a row that does not exist.",
  "honesty_line": "An update is a record of a change, not an announcement of an achievement. Nothing here upgrades a row; the rows say what they prove.",
  "kinds": [
    {
      "id": "added",
      "label": "Row added"
    },
    {
      "id": "corrected",
      "label": "Row corrected"
    },
    {
      "id": "verified",
      "label": "Evidence re-checked"
    },
    {
      "id": "recorded",
      "label": "Recorded"
    }
  ],
  "entries": [
    {
      "date": "2026-09-22",
      "kind": "recorded",
      "row": null,
      "headline": "The mailbox was swept for memberships we hold and were not showing; five were missing and three claims did not survive the sweep",
      "detail": "Every message in nicholas@csoai.org for 2026 was read by header, and the bodies were searched server-side for each body by name. Five things we hold had no row and now have one: Open Invention Network, LOT Network, the EU AI Pact, our comments to BSI committee ART/1, and the mcpservers.org listing. Three did not survive. MLCommons has no membership record at all — the only message naming it is one we wrote about a measurement we published concerning them. OpenML got as far as a verification request and no further. Glama's listings are real but have no public URL a stranger can open. Those three are recorded under Named, and not listed, with the reason, because a membership we cannot show is not the same as one we do not have, and neither may be displayed as current.",
      "evidence": "https://councilof.ai/interop/memberships.json",
      "evidence_kind": "public_url"
    },
    {
      "date": "2026-09-22",
      "kind": "added",
      "row": "open-invention-network",
      "headline": "Open Invention Network lists CSOAI LTD on its public community of licensees",
      "detail": "The OIN 2.0 License Agreement was signed on 2026-08-15 and OIN's public page names us. It is a royalty-free patent cross-licence over the Linux System definition: a patent non-aggression arrangement, not a standards body and not an opinion about anything we measure.",
      "evidence": "https://openinventionnetwork.com/community-of-licensees/",
      "evidence_kind": "public_url"
    },
    {
      "date": "2026-09-22",
      "kind": "added",
      "row": "lot-network",
      "headline": "LOT Network's public member directory lists CSOAI LTD",
      "detail": "LOT Network wrote \"Welcome CSOAI LTD to LOT Network\" on 2026-08-23. Like the OIN row this is a patent arrangement — licence-on-transfer against patent assertion entities — and nothing about AI governance.",
      "evidence": "https://lotnet.com/members/?dir=1&name_directory_startswith=C",
      "evidence_kind": "public_url"
    },
    {
      "date": "2026-09-22",
      "kind": "added",
      "row": "eu-ai-pact",
      "headline": "The Commission's AI Office recorded us as having joined the AI Pact on 2026-08-20",
      "detail": "Its mail opens \"Thank you for having recently joined the AI Pact\". An earlier note of ours had reasoned that the AI Pact probably did not apply to a measurement company; the mailbox says we joined it, so the note was wrong and the row is the correction. We have not signed the voluntary pledges and the Commission's public page does not name us, so the row is private evidence and claims neither.",
      "evidence": "INBOX 7036 (EU-AI-PACT@ec.europa.eu, 2026-08-20) and INBOX 7035 (2026-08-20)",
      "evidence_kind": "private_email"
    },
    {
      "date": "2026-09-22",
      "kind": "added",
      "row": "bsi-art1-comments",
      "headline": "BSI passed fourteen of our comments on a draft British Standard to the manager of committee ART/1",
      "detail": "The comments on draft BS ISO/IEC 4213 (project 2024-01986) were sent on 2026-09-15 and acknowledged the same day. The row says contributor, not member: commenting on a public draft is open to anyone, and we have not been nominated to ART/1.",
      "evidence": "INBOX 8363 (AdminStart@bsigroup.com, 2026-09-15) and INBOX 7116 (2026-08-22)",
      "evidence_kind": "private_email"
    },
    {
      "date": "2026-09-22",
      "kind": "added",
      "row": "mcpservers-org",
      "headline": "The mcpservers.org directory approved the Council of AI GSPC server on 2026-09-05",
      "detail": "Recorded as private evidence rather than a public URL: mcpservers.org answers HTTP 403 to automated clients, and the only place our name appears in that response is the URL we ourselves asked for, echoed back in the challenge. A name that comes from our own request proves nothing about the page.",
      "evidence": "INBOX 7848 (contact@mcpservers.org, 2026-09-05)",
      "evidence_kind": "private_email"
    },
    {
      "date": "2026-09-22",
      "kind": "verified",
      "row": null,
      "headline": "Every public evidence URL in the manifest was re-fetched and answered with our name",
      "detail": "scripts/memberships-check.mjs was run against the live web. Every public-evidence row answered HTTP 200 and named Council of AI, CSOAI or Templeman, except the US Treasury row, which is corrected below. Private-evidence rows were schema-validated and not fetched, because nothing on the network can confirm them; the manifest says so on each one. The check's own --selftest passed first, so the run is one that could have failed.",
      "evidence": "node scripts/memberships-check.mjs",
      "evidence_kind": "command"
    },
    {
      "date": "2026-09-22",
      "kind": "corrected",
      "row": "us-treasury-genius-docket",
      "headline": "Evidence moved from regulations.gov to the Federal Register document, same docket",
      "detail": "regulations.gov answers HTTP 403 to every non-browser client, so the row was failing its own check for a reason that had nothing to do with the filing. The evidence is now Federal Register document 2026-16796, published 2026-08-18, which names docket TREAS-DO-2026-0496 and carries the closing date. The comment is still submitted at regulations.gov and the row still says PENDING: nothing has been filed.",
      "evidence": "https://www.federalregister.gov/documents/2026/08/18/2026-16796/genius-act-regulations-on-payment-stablecoin-issuance-offer-and-sale",
      "evidence_kind": "public_url"
    },
    {
      "date": "2026-09-22",
      "kind": "added",
      "row": "openaire",
      "headline": "OpenAIRE harvested the Zenodo deposit; the record names Council of AI",
      "detail": "A query to OpenAIRE for DOI 10.5281/zenodo.21991105 returns one research product naming Council of AI, CSOAI and Templeman. Harvesting is automatic and the record's own status field read UNDER_CURATION, so the row is indexing, not review.",
      "evidence": "https://explore.openaire.eu/search/publication?pid=10.5281%2Fzenodo.21991105",
      "evidence_kind": "public_url"
    },
    {
      "date": "2026-09-22",
      "kind": "recorded",
      "row": null,
      "headline": "The named-and-not-listed list now carries the bodies we are not eligible for, with the reason",
      "detail": "Absence from the table used to be a gap. It is now an answer. The European Commission's General-Purpose AI Code of Practice is limited to providers of general-purpose AI models and we provide none. OECD.AI's Network of Experts is nominated by countries, not applied to. GPAI membership is held by countries. LF AI & Data's free tier excludes for-profit companies. The IETF has no membership at all. Each of those is recorded with its source rather than left blank.",
      "evidence": "https://councilof.ai/interop/memberships.json",
      "evidence_kind": "public_url"
    },
    {
      "date": "2026-09-22",
      "kind": "added",
      "row": "w3c-agent-conformance",
      "headline": "Joined the W3C Agent Conformance and Benchmarking Community Group on behalf of csoai ltd",
      "detail": "The join page reported that the organisation had already signed the group's Contributor License Agreement, and the public participants page names us the same day. Community Groups are free and open; this is participation, not W3C Membership.",
      "evidence": "https://www.w3.org/community/agent-conformance/participants",
      "evidence_kind": "public_url"
    },
    {
      "date": "2026-09-22",
      "kind": "added",
      "row": "w3c-aikr",
      "headline": "Became a participant in the W3C AI KR Community Group representing csoai ltd",
      "detail": "The W3C Accounts Team recorded the participation at 04:27 UTC. The group has adopted nothing of ours and W3C has reviewed nothing of ours.",
      "evidence": "https://www.w3.org/community/aikr/participants",
      "evidence_kind": "public_url"
    },
    {
      "date": "2026-09-22",
      "kind": "recorded",
      "row": "bank-of-england-code-of-practice",
      "headline": "Consultation response sent to the Bank of England on the systemic stablecoin Code of Practice",
      "detail": "Sent to the address the Bank published for the consultation, with Date and Message-ID headers. The Bank publishes no acknowledgement we can link to, so the row is marked private evidence. A response is a submission, not recognition.",
      "evidence": "Sent 499 (2026-09-22, Message-ID <1509486989.87654.1790069007934@privateemail.com>)",
      "evidence_kind": "private_email"
    },
    {
      "date": "2026-09-21",
      "kind": "added",
      "row": "linux-foundation",
      "headline": "Linux Foundation membership agreement completed by all parties",
      "detail": "DocuSign reported the envelope completed by all parties on 2026-09-21. The member portal record M-018884 (login required) gives the tier as Silver. The public members page did not name us on 2026-09-22, so the row is marked private evidence. Membership is a paid participation tier, not an endorsement of anything we publish.",
      "evidence": "INBOX 8955 (the Linux Foundation, 2026-09-21: one DocuSign envelope carrying the membership agreement for both the Open Secure AI Alliance and the Linux Foundation memberships) and INBOX 8987 (DocuSign, 2026-09-21 22:56 UTC: 'All parties have completed')",
      "evidence_kind": "private_email"
    },
    {
      "date": "2026-09-21",
      "kind": "added",
      "row": "osaia",
      "headline": "Became a member of the Open Secure AI Alliance",
      "detail": "The participation agreement was completed by all parties via DocuSign, and the Alliance's onboarding mail opened 'Thank you for becoming a member', following an application received on 2026-09-18. The Alliance publishes no member list, so the evidence is private. The note CSOAI published to the Alliance's SAFE discussion is our own; we have no record of SAFE adopting or acknowledging it.",
      "evidence": "INBOX 8987 (DocuSign completed, 2026-09-21), INBOX 8988 (2026-09-21) and INBOX 8859 (2026-09-18)",
      "evidence_kind": "private_email"
    },
    {
      "date": "2026-09-21",
      "kind": "recorded",
      "row": "dif",
      "headline": "DIF asked which name to display for our Contributor Member listing; we answered the same day",
      "detail": "The Contributor agreement was completed in the name of CSOAI LTD on 2026-08-18. DIF publishes no member roster we could find — identity.foundation/members answered 404 on 2026-09-22 — so no public page confirms this yet and the row says so.",
      "evidence": "INBOX 8945 (DIF Membership, 2026-09-21) and Sent 487 (2026-09-21)",
      "evidence_kind": "private_email"
    }
  ]
}