AI Policy Wiki
Dashboard
⚲
AI Policy Wiki › Briefings

Gap Scan — 2026-06-11

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

Contents

Scanned

Recent window: 4 dev-log files (2026-06-09-1505, 06-09-2205, 06-10-1513, 06-10-2206), ~50 pages edited in last 48h. Rotation slice: 7 — legislation/ (all 95 pages).

Slice-scan note. The legislation thinness scan initially flagged ~30 pages on sources_count: 0, but inspection showed most flags were a missing sources_count frontmatter field on otherwise well-built, freshly updated pages (EU AI Act (Regulation 2024/1689), America's AI Action Plan, California SB 53, Executive Order 14365, New York RAISE Act) — a lint frontmatter repair, not a content gap. Genuine thin anchors were targeted instead (below).

Gaps actioned (7 of ~22 found)

New pages created (live)

Pages expanded (live)

  • AI LEAD Act (S. 2937) — slice-7 thin anchor (594w × in-degree 62, sources_count absent). Added congress.gov status confirmation (no action since the Sept 29, 2025 Judiciary referral as of June 2026), sponsor committee positions and the FedScoop products-classification framing, and a Reactions section (CCDH support; the OpenAI/SB 3444 contrast). [3 new (Source: URL) cites] sources_count → 3.
  • California SB 243 — Companion Chatbots — slice-7 thin anchor (620w × in-degree 98, sources_count absent; effective since Jan 1, 2026). Added statutory detail from legal analyses: the operator definition and non-delegable compliance, the companion-chatbot definition and express exclusions, the coverage test ("capable of meeting a user's social needs"), detailed disclosure obligations (reasonable-person AI notice; suitability warning; the per-minor three-hour recurring notification and knowledge standard), harm-prevention protocol and minor sexually-explicit-content requirements, annual-report contents, and an Enforcement section ($1,000-per-violation/actual-damages private right of action, cumulative-duties clause, SB 1001 interaction). Also repaired the misleading "Effective July 1, 2027" frontmatter description (core provisions are effective Jan 1, 2026; reporting from Jul 1, 2027). [1 new source, multiply cited] sources_count → 3.
  • Bletchley Declaration (AI Safety Summit, 2023) — slice-7 thin anchor (682w × in-degree 96, sources_count: 1). Replaced the stale forward-looking "AI Impact Summit (Kigali/India, 2026)" reference with the actual fourth summit: India AI Impact Summit 2026, Bharat Mandapam, New Delhi, 16–21 February 2026, concluded with the New Delhi AI Impact Summit Declaration plus the Alliance for Advancing Inclusion Through AI, the Charter for the Democratic Diffusion of AI, and the Equitable AI Transitions Playbook. [2 new (Source: URL) cites] sources_count 1 → 3.

Queued — foundational sources

  • Dario Amodei, "Policy on the AI Exponential" (June 10, 2026) — gap type 1 + live thread (score 8, highest this run): Amodei's move past transparency-first to binding FAA-style regulation — mandatory third-party testing above a compute threshold in four risk areas (cybersecurity, bioweapons, loss of control, automated R&D) with government block/reversal power — plus job-displacement, FDA/EMA-reform, civil-liberties, and democratic-coalition programs; released with an Anthropic legislative proposal and job-displacement framework carrying stated financial backing. Verified: darioamodei.com (canonical author host; full text scraped, distinctive passages confirmed; corroborated by the WSJ-sourced June 10 dev-log and Inside AI Policy). Saved: Raw Sources/Policy on the AI Exponential (Dario Amodei, June 2026).md. Queued: INGEST-amodei-policy-ai-exponential-2026.md. Verification trail: queue/gap-scan/proposed-sources/amodei-policy-ai-exponential-2026.md.
  • Altman & Pachocki, "Built to benefit everyone: our plan" (OpenAI, June 8, 2026) — gap type 1 + live thread (score 8): OpenAI's "third phase" statement — three goals (automated AI researcher with a March-2028 internal belief, accelerating the economy, a personal AGI for everyone) and the endorsement of an international organization able to coordinate "slowing frontier development when needed." The June 10 lint cycle folded the event from secondary coverage; this queues the primary text. Verified: openai.com (canonical corporate host; full text scraped; quoted passages match TNW/AOL and Inside AI Policy coverage verbatim). Saved: Raw Sources/Built to Benefit Everyone — Our Plan (Altman & Pachocki, OpenAI, 8 June 2026).md. Queued: INGEST-openai-built-to-benefit-everyone-2026.md. Verification trail: queue/gap-scan/proposed-sources/openai-built-to-benefit-everyone-2026.md.

Authenticity-verification failures

  • None. Both pulled foundational sources verified cleanly against canonical hosts (darioamodei.com, openai.com); the three expansion source sets resolved to their publishers (congress.gov, fedscoop.com, counterhate.com, gunder.com, indiaai.gov.in, en.wikipedia.org) without conflicts.

Resolved as not-a-gap

  • The \| "broken-link" cluster (15 files, e.g. [[companies/anthropic\]], [[california-sb-243\]]) — false positive. These are wikilinks inside markdown tables using Obsidian's required escaped-pipe alias syntax ([[page\|label]]); they render correctly and must NOT be "fixed" (unescaping would break the tables). The broken-link scanner should learn the \| escape — flagged for lint.
  • Section 1260H list (deferred 06-10) — resolved by the 06-10 lint/ingest cycle, which created Section 1260H List (DoD Chinese Military Companies).

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

  • Munich Regional Court Google AI Overviews ruling (case 26 O 869/26, decided May 28, public June 9 — Google directly liable for AI Overviews' false statements as its own content) — litigation/ page candidate, score 5; sits in the unprocessed 06-10-2206 digest, so the nightly developments-log fold gets first pass; party names not yet confirmed for litigation frontmatter. Re-evaluate next run.
  • Narayanan & Kapoor, "Why AI hasn't replaced software engineers" (June 10) — foundational-essay candidate (the "AI washing" argument; NY WARN-act checkbox data; Fed software-employment estimates), score 5; over cap this run.
  • claude-code (in-degree 8) — perennial; still blocked on the product-page placement decision (carried since 2026-06-04). Score 3. Needs a schema call from the user.
  • Florida ACLU facial-recognition wrongful-arrest suit (Dillon) — single news item, sources conflict on the arrest year; quality-gate deferral. Score 3.
  • Slice-7 thin anchors not reached: Executive Order 14110 (RESCINDED) (944w × 80, src 1), Seoul Frontier AI Safety Commitments (2024) (899w × 97, src 1), G7 Hiroshima Code of Conduct for Advanced AI (2023) (892w × 55, src 1), Paris AI Action Summit Declaration (2025) (941w × 59, src 1), California SB 1047 (VETOED) (1155w × 43, src 1), Singapore Model AI Governance Framework for Generative AI (2024) (738w × 33, src 1), Executive Order 13859 (385w × 19), SCALE Act — Rep. John Moolenaar (April 2026) (606w × 19, src 1), CREATE AI Act (Young) (440w × 6). Score 2–3 each.
  • Lint territory (route to lint, not gap-scan): missing sources_count fields across slice-7 pages (incl. the New York RAISE Act confidence: high + no-sources inconsistency); "Representative Adam Schiff" on AI Literacy should read "Senator" (see the new Adam Schiff); sources/anthropic-pwc-expanded-partnership (6) and sources/ft-mundy-americans-dread-ai-2026 (3) citation-format conversions; the \| scanner-escape fix above; the AI LEAD Act (S. 2937) "Related" self-link ([[ai-lead-act|Premature Antitrust Standards source summary]]) which mislabels or mistargets its source page.
  • claude-sonnet-45 (4) — alias-vs-page decision; concepts/ai-risk-management (3) — likely alias to NIST AI Risk Management Framework 1.0; entities/mike-rounds (now 2+, from lift-ai-act and the new Schiff page). Score 1–2.
  • Anthropic companion artifacts to the Amodei essay (the frontier-model-testing legislative proposal; the job-displacement policy framework) — separate primary documents; queue once canonical URLs are confirmed, likely via the essay's ingest. Score 4.

Failures

  • Dashboard rebuild failed (fifth consecutive run): iCloud sync-conflict * 2 directories inside Wiki/_meta/site/ are locked against deletion from the sandbox, so bin/build-dashboard.py aborts at rmtree. Host-side fix required — see queue/gap-scan/needs-review/2026-06-11-dashboard-rebuild-failed.md. Content graph unaffected.

One-line summary

Queued the week's two defining frontier-lab policy statements (Amodei's binding-regulation essay; the Altman–Pachocki coordination post) as verified primary texts, created the missing Adam Schiff page behind the HALO Act thread, expanded three high-in-degree legislation thin anchors (AI LEAD, SB 243, Bletchley), fixed 6 alias links — and established that the big \| broken-link cluster is correct Obsidian table syntax, not a gap. User review needed on: the two queued ingests, and the recurring host-side dashboard-rebuild block.

Categories: Briefings | briefing | gap-scan