Observed 2026-09-09. Bounded public-source inquiry; candidate findings only. No canonical-case edits, outreach, publication, or paid access. Scope: identifiable carriers, institutional venues, and the difference between receiving a demand and adopting it. Six evidence clusters follow. The result is documented access plus a continuing civil-society workstream, with no binding adoption of the campaign established by the records inspected here.
1. Maria Ressa carried the named campaign into a UN General Assembly speech
U1 — UN-hosted prepared statement, 22 September 2025. [Speech for UN 80th Anniversary General Assembly](http[local research file] Maria A. Ressa. Read all four pages; decisive locator PDF page 3, “Three Solutions,” first solution. Ressa explicitly names the Global Call, says more than 200 people will launch it later that day, and asks governments to define international boundaries for unacceptable AI risks. This is a submitted/prepared speech on the UN's statement service, not a General Assembly resolution or proof every word was delivered. It supports a named public carrier and recipient venue. It does not establish an agreed UN position, a campaign drafting contract, or who obtained her speaking invitation.
Acquisition: web PDF parser failed; ordinary public GET succeeded (HTTP 200), parsed in memory with pypdf. Original-byte SHA-256: add56fc6bb85bf06b39a110e9aa8f3b49064859125287af9378df636a941e471. Bytes were not retained because this lane was restricted to this note.
2. Yoshua Bengio presented the red-lines initiative at the Security Council; the same meeting contained opposition
U2 — UN-hosted automatic transcript, Security Council meeting 10005, 24 September 2025. [Transcript](http[local research file] Scoped read of Bengio's intervention, especially the paragraph beginning “Earlier this week,” and the United States intervention. Bengio refers to the newly supported international red lines and calls for government visibility and control. The US speaker rejects international bodies asserting centralized AI governance. These are evidence of presentation and disagreement, not a Council decision or proof the US statement targeted this campaign specifically.
The service explicitly states its automatic transcripts are not official records or official UN documents. Its speaker labelling contains errors; use the original meeting catalogue to identify Bengio, not an automated label alone. [Official catalogue S/PV.10005](http[local research file] identifies the meeting and Bengio. The English official-record PDF was indexed but not fully acquired: web parser error, ordinary GET HTTP 202 with empty body. The automatic transcript was the bounded substitute. Hash of obtained transcript HTML: 9b6731f529a5f7ab53d673bffccac05409ac2e9db1a4eca7a0db79f51bb1fa34.
3. UNESCO hosts a continuing workstream with campaign organizers, under an express independence disclaimer
U3a — [International AI Red Lines subgroup](http[local research file] current page, inspected 9 September 2026. Full substantive page read. Locators: “Subgroup Activities,” “Subgroup Members,” and final disclaimer. The Future Society and Ada Lovelace Institute are co-leads; CeSIA is a member. The page describes a six-month research process and three workshops: New Delhi 18 February, Paris/UNESCO headquarters 26 February, and an Ada-organized virtual roundtable 16 March 2026. It distinguishes future explainer/discussion-paper outputs from a linked two-page brief. Activities are independent CSO/academic work and do not necessarily express UNESCO, member-state or partner views or commit UNESCO. This is an organizational seat and a work programme, not delegated rulemaking authority. HTML hash: 5f2d5268da959589a8352a9ec105dd614f3314acd443445db6dd39f11073df5c.
**U3b — [Multi-Stakeholder Coordination Group](http[local research file] Scoped read of International AI Red Lines subgroup entries. UNESCO names Cami Rincon of Ada and Tereza Zoumpalova of TFS. The latter's entry identifies her as a Global Call organizer and leader of TFS's red-lines workstream. This connects an identifiable campaign organizer to the network's coordination structure. It does not specify personal veto, budget, or member-state approval rights. HTML hash: 9a142f15220a0ab3a8b3c4dd17ae3bfcf04697f3f573844e3f700b3c1a45630d.
**U3c — [Umbrella network page](http[local research file] Read purpose, chronology, activities and funding paragraphs. The network launched in June 2025, before the September campaign; members shape its activities. UNESCO reports initial Patrick J. McGovern Foundation funding through its 2025 inclusive ethical-governance project, and an earlier UKRI-funded pilot. This concerns the umbrella network. It establishes neither a Global Call grant nor campaign-specific donor control. HTML hash: 532614bd3086d8a9d30339eadb688bf283a979577c435b67e95b2ce42da9b6f8.
Output-access frontier: the subgroup's visible Download PDF link resolves to [UNESCO's SharePoint share](http[local research file] Public link GET returned viewer HTML; the exact file path exposed by that redirect (CSO_Redlines_compressed.pdf) returned HTTP 401. No authentication or bypass attempted; full brief uninspected. Its title/description are website evidence only. Missing decisive record: readable brief plus version/approval record and any transmission to a member-state delegation.
4. The campaign reached a UK parliamentary demand; ControlAI support is separately acknowledged
U4 — [Hansard, 29 January 2026](http[local research file] “Superintelligent AI.” Scoped read: Hunt opening at 3.03 pm, Clement-Jones at 3.42 pm, and opening ministerial response at 3.50 pm. Hunt thanks ControlAI for support. Clement-Jones says he joined the Global Call, promotes it, then separately describes ControlAI's UK proposals. He declares DLA Piper consultancy and APPG AI co-chairing. The minister, Baroness Lloyd of Effra, describes context-based regulation and AISI work without accepting a campaign treaty or moratorium in the inspected passage.
The earned mechanism is acknowledged NGO support and a signatory's explicit parliamentary advocacy. Support is not evidence of a payment, authorship of Hunt's speech, or government commissioning. Clement-Jones's interpretation of the call must not silently replace the campaign's own text. The minister's response is not a comprehensive proof of permanent rejection. Web reader content inspected; native GET returned403, so no original-HTML hash or retained bytes claimed. Needed: the actual briefing sent to Hunt/Clement-Jones, correspondence on requested wording, and any departmental response to the specific campaign proposal.
5. India provides a matched access/outcome comparison
U5a — Organizer's account: [TFS, Key Takeaways from the India AI Impact Summit](http[local research file] February 2026 event; publication date not independently established. Read “Defining and Governing Unacceptable AI Risks” and “General Reflections.” TFS says it and CeSIA co-organized a session within AI Safety Connect Day, explicitly building on the campaign. It reports interventions from Denmark's tech ambassador Anne Marie Engtoft Meldgaard, Brazil foreign-ministry AI adviser Guilherme Fitzgibbon Alves Pereira, Japan AISI's Akiko Murakami, Singapore IMDA's Wan Sie Lee, and MEP Brando Benifei. This is named organizer-reported access, not independently confirmed government endorsement. Its own reflections call for moving beyond voluntary commitments. Web reader inspected; ordinary GET403; no captured-original hash claimed.
**U5b — Government outcome: [PIB/MeitY, New Delhi Declaration](http[local research file] Posted 21 February 2026; live page includes an endorsement list dated 5 March, so it is not an immutable launch-day snapshot. Full declaration annex read, especially “Secure and Trusted AI” and “Conclusion.” It supports industry-led voluntary measures and a voluntary, non-binding Trusted AI Commons; the conclusion likewise emphasizes voluntary/non-binding instruments. The declaration does not name this campaign or enact its requested internationally enforceable prohibitions. No drafting or transmission record links the side event to this text. HTML hash: cf42f235597ee26c30934bd44d5c0e7b5f25dba4737595a7cec02fa09c98cab7.
Needed to move from attendance to uptake: the workshop paper actually transmitted to the Safe and Trusted AI working group, recipient acknowledgment, tracked draft/declaration edits and decision record. Independent government priorities and pre-existing summit commitments remain adequate alternative explanations for shared language.
6. Geneva: do not promote speeches or a summary into a negotiated agreement
U6a — [UN Global Dialogue website](http[local research file] and [FAQ](http[local research file] Indexed official-source text was inspected; direct web and ordinary GET access returned403. The homepage advertises a Co-Chairs' Summary for Geneva 6–7 July 2026. The FAQ says the Dialogue is not a negotiating forum and ends with a co-chair summary. The summary itself was not acquired or read**, so no claim is made about whether it incorporates particular red-lines recommendations.
**U6b — [UN automatic closing-session transcript](http[local research file] Scoped review: final co-chair remarks (parsed lines1038–1053) and red-lines search matches. The co-chairs describe diverse views and a summary still to be drafted. The shared-red-lines passage at parsed lines915-918 is a stakeholder intervention, not a negotiated decision. This ASR source has the same non-official-record limitation as U2. It supplies useful context, not an authoritative final outcome. A separate [UN Web TV programme, 1 July 2026](http[local research file] names Bengio and Ressa as scientific-panel co-chairs and describes independent scientific assessments; it does not give them treaty-making authority. Programme metadata inspected; video not watched.
Priority missing record: actual final Co-Chairs' Summary and annexed submissions, especially any TFS/CeSIA/CHAI contribution, followed by the attribution/version trail. Root should treat this as a defined public-document access gap, not evidence of concealment or adoption.
Candidate typed relations (not yet canonical)
IDs below are provisional within this note; match existing entity IDs before integration. Evidence status describes the source claim, not a finding of control. No graph edge should jump directly from campaign involvement to a state adopting rules.
| Candidate ID | Direction and relation | Date / action stage | Source and limitation |
|---|---|---|---|
| rl-policy-01 | Maria Ressa → Global Call: publicly advocated | 2025-09-22 / prepared statement | U1; named campaign |
| rl-policy-02 | Maria Ressa → UNGA80 anniversary session: presented policy demand | 2025-09-22 / submitted speech | U1; not a resolution |
| rl-policy-03 | Yoshua Bengio → UNSC10005: advocated international AI red lines | 2025-09-24 / speech reported by ASR | U2; official-record corroboration outstanding |
| rl-policy-04 | The Future Society → UNESCO-network red-lines subgroup: co-lead | observed2026-09-09 / current listed role | U3a; independent CSO work |
| rl-policy-05 | Ada Lovelace Institute → same subgroup: co-lead | observed2026-09-09 / current listed role | U3a |
| rl-policy-06 | CeSIA → same subgroup: member | observed2026-09-09 / current listed role | U3a; no control right inferred |
| rl-policy-07 | Tereza Zoumpalova → Global Call: organizer | historical, observed2026-09-09 | U3b; UNESCO-hosted biography |
| rl-policy-08 | Tereza Zoumpalova → UNESCO network coordination group: named subgroup representative | observed2026-09-09 / listed participation | U3b; not UNESCO officer status |
| rl-policy-09 | ControlAI → Lord Hunt: acknowledged support | 2026-01-29 / support reported by recipient | U4; modality/amount/briefing contents unknown |
| rl-policy-10 | Lord Clement-Jones → Global Call: self-reported signatory and parliamentary advocate | 2026-01-29 / delivered speech | U4 |
| rl-policy-11 | TFS and CeSIA → India AI Safety Connect workshop: co-organized | 2026-02 / organizer-reported event | U5a; no state endorsement inferred |
| rl-policy-12 | Named officials in U5a → workshop: interventions reported | 2026-02 / organizer account | U5a; preserve each person's contemporaneous role |
No adoption edge proposed for India, UNESCO, UK Government, UNGA, UNSC or Geneva based on this pass. No campaign-funding edge proposed from the UNESCO umbrella funding paragraph. A current institutional co-chair appointment is a separate governance relation, not evidence the appointment was caused by campaign participation.
Causal checks and stopping boundary
The September2025 campaign cannot have caused the February2025 Paris summit, June2025 UNESCO umbrella-network launch, or earlier Seoul2024 commitments. Its organizers can have contributed to earlier processes in other capacities; that requires its own dated evidence. Hosting on OECD/UNESCO sites is not by itself institutional endorsement. Attendance and parliamentary speech are useful mechanisms of access, but do not establish agreed policy, government sponsorship, donor instruction, or exclusive access.
Closed routes: UN Security Council official PDF202/empty → UN ASR substitute; UN Dialogue homepage/FAQ403 → indexed text and closing ASR only; UNESCO brochure public viewer → exact file401, parked; TFS/Hansard native403 → already-inspected web reader. No continued retries, authenticated sessions, or evasion. An initial Python terminal encoding error occurred after relevant India extraction; it did not invalidate bytes or the web-read annex.
The strongest next acquisition is the actual briefing/transmission document at one conversion point: UNESCO subgroup to a government delegation, TFS/CeSIA workshop to India's working group, or ControlAI to Hunt. These can distinguish policy influence from ordinary participation without enlarging the membership roster.