AI Policy Wiki
Dashboard
⚲
AI Policy Wiki › Briefings

Gap Scan — 2026-07-10

Daily gap hunt — what the wiki is missing and how each gap was triaged.

Contents

Scanned

Recent window: 4 dev-log files (07-08 morning/evening, 07-09 morning/evening; the 07-09 pair is new since yesterday's scan), ~58 pages edited in last 48h. Rotation slice: 8 — litigation/ (27 pages) + standards/ (6) + government/ (19).

Gaps actioned (7 of ~18 found)

New pages created (live)

  • AI Futures Project — live-thread gap: the group published AI 2040: Plan A on July 9 (both 07-09 dev-logs), and despite the wiki holding AI 2027, Daniel Kokotajlo, and URL citations to three of the group's blog posts, the organization itself had no page. Built from ai-2040.com, the group's blog, WaPo/Axios coverage, and the existing ai-2027/kokotajlo pages; first-mention wikilinks added on those two pages so the new page is not an orphan. confidence: medium. [Depends on queued sources: AI 2040: Plan A and the Grading retrospective — cited inline with a provenance note.]

Pages expanded (live)

  • Garcia v. Character Technologies, Inc. — top slice thin anchor: in-degree 71 on sources_count: 2, status last checked 04-28. Added a Settlement section from January 2026 coverage: the "mediated settlement in principle" filing language and stay request, the same-week parallel settlements from Colorado, Texas, and New York families, the undisclosed-terms status, the August 2024 $2.7B Google/Character.AI licensing-deal context, and Character.AI's October 2025 under-18 ban. sources_count 2 → 5; last_status_check → 2026-07-10. [3 new (Source: URL) cites: CNBC, NYT, People.]
  • NYT v. Microsoft, OpenAI et al. — slice thin anchor #2 turned live thread: in-degree 35 on sources_count: 2, status last checked 04-25 — and the 07-09 dev-logs missed that the publisher plaintiffs filed a sanctions motion against OpenAI on July 9 (alleged withholding/destruction of evidence, including tools and datasets that could identify copyrighted journalism in ChatGPT outputs). Added the sanctions-motion section (Reuters, TechCrunch), spring-2026 discovery developments (Brockman diary, Musk-case testimony per Bloomberg Law), the April 2 SJ-briefing close and April 8 30(b)(6) ruling. sources_count 2 → 5; last_status_check → 2026-07-10. [4 new (Source: URL) cites.]
  • None this run — the 07-09 mainspace broken-link scan's remaining high-count targets are all either queued ingests (sources/trustworthy-agents-in-practice, sources/claude-sonnet-5-system-card, sources/how-will-openai-compete-evans — links resolve at ingest) or the curator-blocked product-placement cluster (claude-code, claude-cowork).

Queued — foundational sources

  • AI 2040: Plan A (AI Futures Project, July 9, 2026) — live-thread foundational work, successor scenario to AI 2027. Verified: ai-2040.com (the org's own site), corroborated by WaPo, Axios, ACX. Queued: INGEST-ai-2040-plan-a-2026-07-10.md (URL-only convention — large multi-chapter scenario plus supplements).
  • "Grading AI 2027's 2025 Predictions" (Lifland & Kokotajlo, Feb 12, 2026) — dangling foundational reference deferred since 07-08; cited by URL from Overview and Daniel Kokotajlo, broken-linked from Situational Awareness: A One-Year Retrospective — Nathan Delisle (LessWrong, June 2025). Verified: blog.aifutures.org. Full text saved: Raw Sources/Grading AI 2027s 2025 Predictions.md. Queued: INGEST-grading-ai-2027-2025-predictions-2026-07-10.md. Ingest note: it is the group's self-assessment — Daniel Kokotajlo currently calls it "third-party" and should be corrected at fold.
  • EDPB Guidelines 02/2026 on Anonymisation — companion primary text to the queued web-scraping guidelines, adopted at the same July 8 plenary; deferred yesterday (score ~3). Verified: edpb.europa.eu PDF + consultation page (comments to Oct 30). Queued: INGEST-edpb-anonymisation-guidelines-2026-07-10.md.

Authenticity-verification failures

None. All three queued sources verified on primary hosts (ai-2040.com, blog.aifutures.org, edpb.europa.eu).

Needs review

  • [[concepts/inverse-cooking-problem]] / [[concepts/inverse-trust-problem]] — the 07-03 create-candidates fail the attribution check: the terms appear nowhere in Raw Sources/, the dev-logs, or any other mainspace page — they are unattributed coinages on AI Content Saturation (\"AI Slop\") and AI Fluency Divide, a house-style violation the quality gate declines to compound with new pages. Note with two resolution options left at Wiki/_meta/queue/gap-scan/needs-review/2026-07-10-inverse-cooking-trust-coined-terms.md.
  • Product-page placement (carried) — addendum added to the 07-09 note: ChatGPT Work (July 9 launch) and Cursor "Sand" (reported July 9) now pose the same schema question as claude-code/claude-cowork; curator decision still pending.

Deferred backlog (over the daily cap — re-surfaces next run)

  • China's July 2026 AI-agent and AI-ethics instruments — the IAPP analysis (07-09 evening dev-log) names three July regulatory developments; China — Interim Measures for the Administration of AI Anthropomorphic Interaction Services covers one, but the AI-agent security and ethics measures need primary-source identification (CAC/TC260 documents) before pages can be built (score ~4).
  • Slice thin anchors below the top two: Mobley v. Workday, Inc. (in-deg 25, src 2), DHS — Department of Homeland Security (AI Deployer) (in-deg 16, 385 words), Defense Innovation Unit (DIU) (in-deg 15, 428 words), Department of Energy (DOE) (in-deg 9, 248 words, src 1) (score ~2–3 each; two-per-run pacing).
  • New companies from the 07-09 dev-logs: PrismML (on-device compression), Orbital Compute (orbital data centers), Lancium (Stargate power developer) — each single-source or paywalled-source; wait for corroboration (score ~3).
  • Kokotajlo retrospective link grading-ai-2027-2025-predictions (2 inbound) — resolves when the queued ingest creates the sources/ page; no action needed beyond the queue.
  • California AI safety commission bill — still needs bill-number identification (score ~3, carried since 07-08).
  • Pam Bondi (2 inbound from President's Council of Advisors on Science and Technology (PCAST)), concepts/ai-coarse-grainings (2 inbound), concepts/foundation-models (2 inbound) — standing low-priority backlog (score ~1–2).
  • Duplicate entities/cdao / government/cdao (slice pages, in-deg 56 combined) — lint territory, carried again.

One-line summary

One live page (the AI Futures Project finally has a home), the litigation slice's two biggest thin anchors reinforced and status-checked — catching a July 9 sanctions motion against OpenAI the dev-logs missed — three primary sources verified and queued, and the inverse-cooking/inverse-trust coinages flagged for cleanup rather than page creation.

Categories: Briefings | briefing | gap-scan