AI Policy Wiki
Dashboard
⚲
AI Policy Wiki › Briefings

Gap Scan — 2026-07-14

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

Contents

Scanned

Recent window: 4 dev-log files (07-12 morning/evening, 07-13 morning/evening — the 07-13 evening file is unprocessed and left for tonight's developments-log cycle; only gap-relevant items actioned), ~49 pages edited in last 48h. Rotation slice: 12 — sources/ I–P (~96 pages).

Gaps actioned (8 of ~18 found)

New pages created (live)

  • None this run — the day's top gaps were queue-lane foundational sources (three matured carry candidates plus the RSP v3.4 primary text) and slice thin anchors, not missing pages. The one new-page candidate in the window (the Stanford Digital Economy Lab "We Must Act Now" statement, 07-13 evening) belongs to tonight's fold first (deferred below).

Pages expanded (live)

  • Helsing — the 07-13 lint report's explicit gap-identifier handoff: single-source (CNBC), confidence: low, founding facts from general knowledge. Added founders (Reil, Scherf, Köhler; March 2021, Munich), the full product line (HF-1/HX-2 GPS-independent strike drones, Altra recce-strike platform, Cirra electronic warfare on Eurofighter ECR, SG-1 Fathom + Lura underwater systems, CA-1 Europa with 2027 first-flight target, €580M CFSN contract, June 2025 Saab cooperation), Ukraine deployments (4,000 Germany-funded HF-1s; 6,000 HX-2s agreed early 2025; Bloomberg-reported suspension of new HX-2 orders over battlefield performance), funding history (Ek/Prima Materia €100M Nov 2021 with the Spotify-artist boycott episode; €209M Series B Sept 2023; €450M at ~€5B July 2024; €600M at €12B June 2025; shareholder stakes), and the 07-13 evening dev-log's Series E detail ($18B valuation; Goldman Sachs Alternatives, Dragoneer, Iconiq, CPPIB, JPMorgan), with a new Snapshot valuation table. Enrichment provenance note retired as fulfilled. confidence low → medium; sources_count 1 → 4. [4 new (Source: URL) cites]
  • NIST AI Risk Management Framework (AI RMF 1.0) — the slice's most-relied thin anchor: in-degree 130 (highest in sources/ I–P) at 629 words, confidence: medium, no sources_count, stale since 06-06. Corroborated against the NIST primary host: DOI 10.6028/NIST.AI.100-1 and canonical PDF, the 2021–2022 consensus development process (RFI, concept paper, two public drafts, workshops), the Playbook/Roadmap/Crosswalk/Perspectives companion set, the AIRC launch (March 30, 2023), official Arabic and Japanese translations, and NIST's statement that "the AI RMF 1.0 is being revised." Frontmatter repaired (source_url, sources_count: 3). Every prior fact and citation preserved. [2 new (Source: URL) cites]
  • OpenAI Preparedness Framework V.2 — slice thin anchor #2 (in-degree 107, high on sources_count: 1 — the stochastic-parrots pattern). Corroboration against OpenAI's own launch announcement surfaced a factual error: the page presented a four-level Low/Medium/High/Critical table as the v.2 structure, while the primary host states v.2 "streamlined levels to two clear thresholds" (High / Critical). Corrected with an inline correction note. Added the five tracking criteria (plausible, measurable, severe, net new, instantaneous or irremediable), Research Category focus areas (Long-range Autonomy, Sandbagging, Autonomous Replication and Adaptation, Undermining Safeguards, Nuclear and Radiological), the persuasion-out-of-scope provision, Capabilities/Safeguards Reports, the frontier-landscape adjustment clause, and scalable-evaluation methodology. source_url added; sources_count 1 → 2. [5 new (Source: URL) cites]
  • AI-Enabled Terrorism — CASP name corrected: the page (created 07-13 from dev-log material) expanded CASP as "Centre for the Study of Armed Groups and Proliferation"; the organization's own site gives "Cambridge Programme on AI Science & Policy." Added the report's full title ("'God has helped us, and so will AI': How the Terrorist Group Boko Haram Uses Frontier AI"), author (Antonia Juelich), and abstract-corroborated detail (northeast Nigeria interviews; activity primarily through 2024; specialized units; safeguard circumvention; openness to mass-casualty weapons though documented use remains conventional). sources_count 2 → 3.

Queued — foundational sources

  • Anthropic Responsible Scaling Policy v3.4 (effective July 8, 2026) — type-1 dangling foundational reference on a live thread and a 07-13 lint "suggested next source": the series is tracked by name (Anthropic's Responsible Scaling Policy (Version 2.2), Anthropic's Responsible Scaling Policy (Version 3.1)) but the v3.4 deltas folded 07-13 onto Responsible Scaling Policy (RSP) are cited only inline. Verified: anthropic.com landing page changelog matches all five wiki-recorded deltas; canonical PDF + redline URLs captured. Queued URL-only: INGEST-anthropic-rsp-v34-2026-07-14.md.
  • Nathan Lambert, "6 months to live for open models" (Interconnects, July 12, 2026) — matured 07-13 carry: the nightly fold has occurred and Open-Weight Frontier Models + Overview now summarize the essay by name; the 07-13 evening dev-log adds the WaPo open-source-framework exclusive and Bloomberg distillation-debate items to the same thread. Verified on the author's own publication; final section is subscriber-only, so no raw was saved (excerpt rule) — queued URL-only with the paywall noted: INGEST-lambert-6-months-open-models-2026-07-14.md.
  • Benedict Evans, "Ways to think about token pricing" (July 9, 2026) — matured 07-13 carry: Inference Economics and Token Pricing names and summarizes it. Verified on ben-evans.com; full text saved: Raw Sources/Ways to Think About Token Pricing.md. Queued: INGEST-evans-token-pricing-2026-07-14.md.
  • CASP, "'God has helped us, and so will AI': How the Terrorist Group Boko Haram Uses Frontier AI" (Antonia Juelich, 2026) — type-1 dangling foundational reference anchoring AI-Enabled Terrorism; a 07-13 lint suggested source. Verified on casp.ac (abstract page; title/author/findings corroborated against the NYT investigation and dev-logs); full report document to be located at ingest. Queued URL-only: INGEST-casp-boko-haram-frontier-ai-2026-07-14.md.

Authenticity-verification failures

  • None. All four queued sources verified on their canonical hosts (anthropic.com, interconnects.ai, ben-evans.com, casp.ac). Two provenance cautions recorded rather than failures: the Lambert essay's final section is paywalled (raw not saved; URL-only queue), and the casp.ac fetch is abstract-only with the site served via a lovable.app deployment behind the programme's own domain (full document to be located at ingest).

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

  • "We Must Act Now" 88-word statement (Stanford Digital Economy Lab, July 13; 200+ signatories incl. Stiglitz, Acemoglu, Johnson, Schmidt, Hoffman, Dean, Clark, Friar) — substantive public statement, foundational queue candidate once tonight's cycle folds it and a wiki page names it (Bletchley/CAIS-statement pattern). Score ~4.
  • H.R. 9619 (People-First Chatbot Act) primary text — re-checked today: Congress.gov still shows "text has not been received" as of 07/14. Carried; re-check next run. Score ~3.
  • Slice-12 residue (two-per-run pacing): Paris AI Action Summit Declaration (2025) (in-deg 61, high/1), OpenAI Model Spec (370 words at in-deg 26), Project Glasswing: Securing Critical Software for the AI Era (54, no sources_count), On the Biology of a Large Language Model (45), Machines of Loving Grace (44), International AI Safety Report 2025 (29, high/1), New York RAISE Act (S. 8828) (in-deg 107 but 2,273 words — frontmatter repair only). Score ~3–4 each.
  • US AI Regulatory Approaches Compared (in-deg 36) and Labor Disruption Timelines: Who Predicts What and Why (in-deg 21) — carried from 07-12/07-13; need a run with capacity. Score ~4 each.
  • Pangram Labs social-media study — primary figures already folded onto AI Content Saturation (\"AI Slop\"); borderline foundational. Score ~3.
  • Possible standalone models/gemini-3-5 page once 3.5 Pro ships (reported targeting July 17). Score ~3.
  • Xi Jinping's first World AI Conference keynote (July 17, Shanghai; 07-13 evening dev-log) — fold is tonight's cycle; a summit-outcomes thread may warrant page work after the event. Score ~2.
  • Standing carries: claude-code/claude-cowork placement (curator-blocked, needs-review note of 07-09/07-10), inverse-cooking/inverse-trust coined terms (needs-review 07-10), duplicate entities/cdao/government/cdao (lint territory), concepts/foundation-models disambiguation, person/org pages at in-degree 2 (pam-bondi, masahiro-mori, anil-seth, eric-horvitz, orin-kerr, alondra-nelson, palmer-luckey, tom-mitchell, soufan-center), Helix (Figure AI Vision-Language-Action model) (next models slice). Score ~2–3.

One-line summary

Four foundational sources verified and queued for review (RSP v3.4, the Lambert open-models essay, the Evans token-pricing essay, the CASP Boko Haram report — the first two anchoring the week's open-weight-EO policy thread); three thin anchors expanded live (helsing enrichment per the lint handoff, and the slice's two most-linked source pages, nist-ai-rmf and openai-preparedness-framework — the latter correcting a four-level-threshold mischaracterization against the primary host); one naming error fixed (CASP). Nothing failed verification; the queue awaits the user's ingest review.

Categories: Briefings | briefing | gap-scan