Gap Scan — 2026-05-29

Live run. Seven new pages were researched and committed directly to their real Wiki/ folders (status: active; all confidence: medium). Two thin anchors in today's rotation slice (slice 8 — litigation/ + standards/ + government/) were expanded with full-section build-outs. One foundational primary text — the FBI IC3 2025 Internet Crime Report — was authenticity-verified against ic3.gov and queued for ingest (pull-at-ingest, per skill protocol for large federal PDFs). Three link aliases ([[legislation/eo-trump-state-ai-preemption-draft]] → [[legislation/eo-trump-december-2025-state-preemption]]) were fixed across two pages. No authenticity-verification failures.

Scanned

Recent window (last 48h): 4 New Developments Log/ files (2026-05-27-1505, 2026-05-28-0505, 2026-05-28-1110, 2026-05-28-2205); ~78 wiki pages with last_updated inside the window. Dominant clusters: the state-AI-bill triple-finish (IL SB 315 passed + Pritzker pending; CT SB 4 + SB 5 signed; NY Safe By Design Act included in FY27 budget) all on May 28; the Anthropic Series H + Opus 4.8 double; the AI-cost-rationing counter-narrative (WSJ "Corporate America rations AI"); the UK ControlAI ASI Security Bill + Stephen Fry endorsement; the Cognition $26B / AWS-Snowflake $6B / AWS Alexa-for-Shopping / Headway facial scans / AlphaProof Nexus 9 Erdős proofs / DOJ Polymarket insider-trading items.

Rotation slice: 8 of 14 — litigation/ (17 pages) + standards/ (6 pages) + government/ (14 pages).

Broken-link analysis: 574 unique broken targets (essentially flat against yesterday's 562). After alias resolution, doc-artifact filtering, and dropping items already actioned in the May 22 → 28 gap-scan reports — and items with an existing Raw Sources/ file or queued INGEST- task — ~24 genuine new gaps remained.

The dominant cluster today: the May 28 state-AI-bill triple (CT SB 4 + SB 5; NY Safe By Design Act; IL SB 315 Pritzker pending) where the same 24-hour cycle put three Democratic-trifecta states in finish-line motion on overlapping-but-distinct AI architectures (frontier reporting vs. ADMT vs. AI-companion regulation vs. data-broker registration). The UK ASI Security Bill + Fry endorsement is the second cluster, tying together ControlAI (new), Sir Stephen Fry (new), and UK Artificial Superintelligence Security Bill (ControlAI draft, May 28, 2026) (new) — the first major-jurisdiction draft bill to criminalize superintelligence development. The AWS-Snowflake $6B Graviton agentic-CPU deal is the third — anchoring a CPU-side counterweight to the GPU-dominated AI Infrastructure Capex story. The rotation slice surfaced NIST AI Risk Management Framework 1.0 (213 inbound × 384 words — the wiki's second-highest in-degree thin anchor after RSP) and CISA — Cybersecurity and Infrastructure Security Agency (AI Deployer) (296 words × 44 inbound, with significant May 2026 operational updates pending) as the priority expansion targets.

Gaps actioned (10 of ~24 found)

New pages created (live)

• Connecticut SB 5 — Broad AI law (frontier reporting + ADMT + AI companions + sandbox) — gap type 3 + live thread + dangling legislation reference: Connecticut's broad AI law signed by Gov. Lamont May 28, 2026 after a 2024 veto-threat round. Three-year compromise carrying four distinct regimes (frontier-model anonymous risk reporting; ADMT disclosure for consequential employment decisions; AI-companion rules including a flat under-18 ban; regulatory sandbox by Jan 1, 2028). Built with the architectural framing — Maroney's "bundle four regimes in one act" approach as the structural choice — and cross-linked to IL SB 315 (frontier-reporting thresholds), CA SB 243 (companion-AI restrictions), and the simultaneous NY Safe By Design Act. 2 sources (IAPP write-up + the CGA primary text URL); confidence: medium.

• Connecticut SB 4 — Data-broker registration + geolocation-sales ban + facial recognition (CTDPA amendment) — gap type 3 + live thread + dangling legislation reference: Connecticut's privacy half of the May 28 Lamont package — Delete-Act-style data-broker registration (effective Jan 1, 2027), geolocation-data-sales ban, facial-recognition provisions amending the CTDPA, and a streamlined single-opt-out / deletion mechanism due July 1, 2028. Cross-walked to California's Delete Act precedent (DROP platform, 285,000+ CA-resident users as of May 2026). 2 sources; confidence: medium.

• New York Safe By Design Act (FY27 budget, May 28, 2026) — gap type 3 + live thread + dangling legislation reference: New York's children's online-safety law included in Hochul's FY27 enacted budget agreement May 28, 2026. Covers social-media, gaming, messaging; restricts adult-to-minor contact; requires parental consent under 13; default design protections for under-17s with no age verification; AI-companion access defaulted off for minors. Page also captures the same-day 43-state-AG opposition to the federal KIDS Act (led by Letitia James), which is the most structurally important AI-federalism move of the week. 3 sources (IAPP + NY Budget Division + NAAG letter); confidence: medium.

• UK Artificial Superintelligence Security Bill (ControlAI draft, May 28, 2026) — gap type 3 + live thread + dangling legislation reference: ControlAI's UK Artificial Superintelligence Security Bill, published May 28, 2026 — the first major-jurisdiction draft to criminalize the development of artificial superintelligence, with personal liability for chief executives and a treaty obligation that the UK pursue an international ban. Built with the contested-context section preserving the "regulate-but-do-not-ban" middle position (EU AI Act / NIST AI RMF / CA SB 53 / IL SB 315 / NY RAISE Act) the bill sits in tension with. 2 sources (ControlAI Substack + ukbill.controlai.com bill landing page); confidence: medium.

• ControlAI — gap type 3 + live thread: UK-based AI-policy advocacy organization (Andrea Miotti CEO, Tolga Bilge co-author, Connor Leahy US Director, Sophie Toura international). Author of the ASI Security Bill and of the October 2025 Call to Ban Superintelligence signed by 100+ UK politicians and a broad swath of AI scientists. Page anchors on ControlAI's single-issue, operationally-legislative position in the AI-policy ecosystem (drafts statutory text and runs constituent-contact campaigns rather than publishing white papers). 2 sources; confidence: medium.

• Sir Stephen Fry — gap type 3 + live thread: British actor, author, broadcaster, and public intellectual. Page is anchored on his May 28, 2026 endorsement of ControlAI's UK ASI Security Bill — the first widely-recognized mainstream cultural-figure endorsement of a development-prohibition AI-policy position the wiki tracks. Full statement reproduced; nuclear-analogy framing flagged as operative (not rhetorical) because it explicitly draws on the UK's history with nuclear non-proliferation, the chemical-weapons ban, and Bletchley. 1 source; confidence: medium.

• Snowflake — gap type 3 + live thread: NYSE-listed data-cloud company (NYSE: SNOW). Built around the May 27, 2026 five-year, $6B AWS Graviton CPU deal for agentic-AI workloads — joining Apple and Meta as one of AWS's largest CPU customers; the deal is the first $6B-scale enterprise contract anchored specifically on agent-orchestration economics (CPU, not GPU) rather than model-serving economics. Page also captures the May 28 semantic-data-layer fight with Microsoft, Databricks, and SAP (Lowe's customer reference as the May 28 datapoint). 2 sources (WSJ + The Information); confidence: medium. Quantitative snapshot block flagged as gap (10-K / Q1 FY27 earnings ingest needed for revenue + customer-count + market-cap).

Pages expanded (live)

• NIST AI Risk Management Framework 1.0 — thin anchor: 213 inbound × 384 words, confidence: high, sources_count: 4 (the wiki's second-highest in-degree thin anchor after RSP, and a slice-8 catch). Expanded to ~1,560 words. Added: the statutory authority (National AI Initiative Act of 2020, 15 U.S.C. § 9401 et seq.), the full citation (NIST AI 100-1, Elham Tabassi 2023, DOI 10.6028/NIST.AI.100-1), the open-and-consensus drafting process (two public drafts, 240+ comments), the explicit CSF structural analogy as why NIST chose the four-function iterative-loop architecture, the AI RMF Playbook / Roadmap / Crosswalks ecosystem, the OMB M-24-10 → OMB M-25-21/22 federal-procurement anchoring (Biden → Trump continuity), the Colorado-AI-Act "reasonable care" link, the Illinois SB 315 third-party audit anticipation, the lab-side RSP / Preparedness inheritance, and the international (OECD, AISI, EU harmonized standards) crosswalks. New cite to the National AI Initiative Act. Bumped sources_count 4 → 5; held at confidence: high.

• CISA — Cybersecurity and Infrastructure Security Agency (AI Deployer) — slice 8 thin anchor: 296 words × 44 inbound, confidence: high, sources_count: 4. Expanded to ~1,160 words. Added: institutional framing (DHS sub-agency, 2018 act, 6 U.S.C. § 651 et seq., Acting Director Gottumukkala, JCDC + KEV/FCEB + 16 critical-infrastructure-sectors role), the operational-role-across-AI-threads section (deployer / defender / international coordinator), explicit framing of the 3-day patch deadline as an 80% compression from BOD 22-01's 15-day standard, the New York State Department of Financial Services (NYDFS) May 26 advisory and the FBI IC3 2025 figures as the broader federal-and-state response wave that CISA sits inside, and an open-questions section (whether the patch deadline is formally adopted; whether the permanent Director nomination is advanced; whether the Grassley letter triggers special-access-exception architecture revision). Bumped sources_count 4 → 5; held at confidence: high.

Link aliases fixed

• [[legislation/eo-trump-state-ai-preemption-draft]] → [[legislation/eo-trump-december-2025-state-preemption]] — 3 instances fixed across 2 pages (Wiki/concepts/ai-personhood.md, Wiki/legislation/missouri-sb-1012.md). The canonical page exists; the broken target was a stub forward-reference left behind from the May 28 Missouri SB 1012 / AI-personhood build.

Queued — foundational sources

• FBI Internet Crime Complaint Center (IC3) 2025 Annual Report — the FBI's 25th-anniversary IC3 annual report containing the AI-attributable cybercrime statistics already folded into the wiki via Fast Company secondary coverage (22,364 AI-related complaints; $893M reported losses; total IC3 complaints surpassing 1M for the first time; total losses topping $20B). Verified: canonical host ic3.gov (FBI's official IC3 domain), URL pattern matches prior verified annual reports (/AnnualReport/Reports/YYYY_IC3Report.pdf), corroborated against the IC3 annual-report index page and Fast Company's May 27, 2026 secondary coverage; distinctive 25-year-anniversary passage matches between firecrawl-extracted preview and secondary press framing. Pull strategy: fetch fresh at ingest time per the skill's large-PDF protocol; verification trail recorded in Wiki/queue/gap-scan/proposed-sources/fbi-ic3-2025-annual-report.md. Queued: Wiki/queue/INGEST-fbi-ic3-2025-annual-report.md. Will become the wiki's primary federal-data anchor for AI and Cybersecurity, with subsidiary anchoring on AI and Misinformation and CISA — Cybersecurity and Infrastructure Security Agency (AI Deployer) (the federal-cyber-deployer page already updated today to flag the broader frontier-AI-cyber-risk surface).

Authenticity-verification failures

None. The FBI IC3 report verified cleanly: ic3.gov canonical FBI-operated .gov domain, URL pattern matches all prior verified annual reports, corroboration via the index page and via independent Fast Company coverage, distinctive-passage match (25-year-anniversary framing appears in both firecrawl preview and secondary press). No signs of fabrication, no satirical / spoofed sources surfaced.

Notes for the next lint / ingest pass

• The May 28 state-AI-bill triple-finish creates a new wiki cross-cutting cluster. Connecticut SB 5 — Broad AI law (frontier reporting + ADMT + AI companions + sandbox), Connecticut SB 4 — Data-broker registration + geolocation-sales ban + facial recognition (CTDPA amendment), New York Safe By Design Act (FY27 budget, May 28, 2026), and the Pritzker-pending Illinois SB 315 (frontier safety framework with mandatory third-party audits) (already updated by today's ingest cycle) are all from May 28, 2026 and represent four different architectures for state-level AI/privacy regulation. A concepts/may-2026-state-ai-finish-line analysis page or — more durably — an expansion of State-Level AI Regulation / AI Federalism to anchor this moment would be a strong next-cycle build.
• legislation/illinois-sb-315 Pritzker-signing update. Today's developments-log fold added the May 28 update (Pritzker signature pending; the $1B catastrophic-risk threshold definition; OpenAI + Anthropic support) — but the page itself still carries the May 16 last_updated. The page should be bumped on its next edit cycle; not a gap-identifier task.
• The 43-state-AG letter (KIDS Act opposition) is currently anchored only on the New York Safe By Design Act (FY27 budget, May 28, 2026) page. When two more state-AG-coalition references accrete, this earns its own entities/state-ag-bloc-on-ai or concepts/state-ag-coalitions-on-ai page.
• entities/andrew-gournardes is referenced from the new NY Safe By Design Act page; he is the bill's sponsor and is also tracked elsewhere in the wiki (per yesterday's alias-check; partial match entities/andrew-gournardes*partial). If that page exists, the new NY page should be re-linked; if not, it's a deferred entity-page candidate (low score — 2 inbound).
• legislation/eu-ai-act is missing from the wiki (verified during NIST AI RMF + ASI Security Bill build-out; the link resolves to no page). Multiple new pages link to it; this is a high-priority deferral for next gap-scan, particularly given the rotation will hit legislation/ again in slice 7.
• entities/letitia-james is referenced from the new NY Safe By Design page (lead of the 43-AG coalition opposing the KIDS Act). 2+ inbound now; promotion-eligible.
• The Anthropic Series H + Opus 4.8 cluster from the May 28 22:05 dev-log file (Anthropic $65B Series H at $965B valuation briefly overtaking OpenAI; Opus 4.8 release; Dynamic Workflows research preview) is the highest-velocity live thread in the wiki right now but is dev-log-skill territory, not gap-identifier territory. Flagging for the next nightly developments-log cycle: Anthropic needs a substantial section update; AI Models needs a claude-opus-4-8 page; the Snapshot Valuation row needs updating.
• concepts/ai-roi-rationalization (the WSJ "Corporate America is starting to ration AI" cluster — Uber, Meta, Microsoft, Salesforce, DoorDash technologists publicly rationing AI use; Marcus on the FT projected negative best-case ROI for Microsoft/Google/Meta/Oracle) is a candidate concept page — strong live thread, but quality-gate-ambiguous (concept page vs. extension of AI Bubble Debate). Score 4, deferred.
• legislation/eu-ai-act page status. During the NIST AI RMF expansion, the EU AI Act link was used as [[legislation/eu-ai-act]] — needs verification that this is the canonical slug; if a different slug exists, this is an alias-fix candidate for next run.

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

• Aei — 4+ inbound after May 27 AEI antitrust-exemption ask; flagged for promotion 2026-05-27 lint. Score 5.
• New York State Department of Financial Services (NYDFS) — 2 substantive inbound (May 21 Asrow letter + May 26 Advisory + today's anchoring from the new CISA — Cybersecurity and Infrastructure Security Agency (AI Deployer) page); flagged for promotion 2026-05-27 lint. Score 4.
• Casey Newton — Platformer journalist; lint-flagged 2026-05-27. Score 3.
• Maxim Topaz — 4 inbound (now actually 5 with today's new pages); three-times-deferred (2026-05-26 → -27 → -28). Score 4. Quality-gate concern (sparse public record outside the Lancet paper) — will become routine once the queued Lancet paper is ingested.
• Seeing Like a State (applied to AI) — Farrell-Shalizi referenced, 6 inbound. Deferred from 2026-05-26 and -28. Score 4.
• AI Political Economy — Farrell-Shalizi cluster, 5 inbound. Deferred from 2026-05-26 and -28. Score 4.
• AI and Children — child-safety topical hole; now 6+ inbound after today's NY Safe By Design Act build. Score 5. Strong promote-next-run candidate.
• Autonomy Certificates — 5 inbound. Deferred from 2026-05-26 and -28. Score 4.
• Jack Dorsey / Brian Armstrong — manager-purge cluster from yesterday's Intelligence Replaces Hierarchy build; 5 inbound each. Score 3.
• Joe Nicola — Missouri SB 1012 sponsor, lint-flagged 2026-05-28 as a stub forward-reference. Score 2; promote on second substantive citation outside the SB 1012 thread.
• Persona — biometric identity-verification vendor (Headway thread). Lint-flagged 2026-05-28 stub. Score 2.
• AI Federalism — 5 inbound stub forward-reference; today's NY / CT / IL / 43-AG cluster strongly motivates this build. Score 5. Strong promote-next-run candidate.
• Andrew Bosworth — Meta CTO, AI-first reorg. Score 3.
• AI Psychosis — promoted from stub today as the rule-of-two threshold accrues (Levie's coinage + WSJ "ration" cluster adjacent framing). Now 4+ inbound. Score 4. Promote on next-cycle when a second AI-bull voice echoes the framing.
• Headway / Lowes / Trajectory / Persona — new May 28 companies from dev-log fold. Score 2 each; defer until a second substantive citation accretes.
• Ai Roi Rationalization — WSJ "Corporate America rations AI" + Marcus negative-ROI cluster. Concept page vs. AI Bubble Debate extension question. Score 4.
• legislation/eu-ai-act — alias-or-missing verification needed; if missing, this is a foundational gap (referenced from dozens of pages including today's expansions). Score 5+, but defer until alias question resolved.
• models/claude-opus-4-8 — new Anthropic model from May 28; dev-log-skill territory primarily but a models/ page is needed. Score 4; flagged for next nightly ingest.
• Slice 8 thin anchors not actioned this run: Chief Digital and AI Office (CDAO) (470w × 65 in-degree, sources_count: 1); DHS — Department of Homeland Security (AI Deployer) (272w × 30 in-degree); GSA — General Services Administration (AI Deployer) (375w × 20 in-degree); NIST AI Agent Standards Initiative (2026) (747w × 55 in-degree, sources_count: 1 — would benefit from a second source). Score 3 / 3 / 2 / 3.
• Finance-firm entity cluster (Kkr, Blackstone, Blackrock, Apollo, Calpine, Sandisk, Eaton, General Catalyst) — pending CLAUDE.md entity_type resolution. Score 2 each, deferred (carried from 2026-05-28).
• OpenAI Strategic Compute Reserve letter (full text) — lint-flagged 2026-05-28 as foundational ingest candidate. Needs House/Senate publication channel to surface the letter (vs. the Substack excerpt). Score 4 (defer until primary text is locatable).
• Greenblatt / Redwood Research "Full automation of AI R&D" post (PDF or full markdown) — lint-flagged 2026-05-28. Score 3.
• AlphaProof Nexus arxiv preprint (arxiv.org/abs/2605.22763v1) — lint-flagged 2026-05-28. Score 3.

One-line summary

Ten substantive actions today: seven new pages live — the May 28 state-AI-bill triple-finish (Connecticut SB 5 — Broad AI law (frontier reporting + ADMT + AI companions + sandbox), Connecticut SB 4 — Data-broker registration + geolocation-sales ban + facial recognition (CTDPA amendment), New York Safe By Design Act (FY27 budget, May 28, 2026)), the UK UK Artificial Superintelligence Security Bill (ControlAI draft, May 28, 2026) + ControlAI + Sir Stephen Fry cluster, and Snowflake anchoring the AWS-Graviton agentic-CPU thread — plus the two slice-8 thin-anchor expansions (NIST AI Risk Management Framework 1.0 384 → ~1,560 words; CISA — Cybersecurity and Infrastructure Security Agency (AI Deployer) 296 → ~1,160 words), three link-alias fixes (eo-trump-state-ai-preemption-draft → eo-trump-december-2025-state-preemption), and one foundational primary text authenticity-verified against ic3.gov and queued for ingest (FBI IC3 2025 Internet Crime Report). No authenticity-verification failures. The strongest deferred candidates for next run are concepts/ai-federalism (5 inbound stub; today's NY/CT/IL/43-AG cluster motivates it) and concepts/ai-and-children (6+ inbound after today's NY Safe By Design Act build).