AI Policy Wiki
Dashboard
⚲
AI Policy Wiki › Analysis & Trackers

Missing Coverage — Residual (post-implementation, 2026-04-14)

high confidence · updated 2026-06-06

What's left after the missing-coverage-2026-04-14 implementation pass that created ~135 pages.

Contents

This page records the state of coverage after the bulk-coverage implementation pass triggered by Missing Coverage: Companies, People, Models, Legislation, Concepts (revised), which created roughly 135 new pages across people, companies, models, legislation, concepts, and think tanks. All ten priority items from that page's "Revised priority order" were addressed at the page-creation level. What remains is page depth, lint hygiene, and a short residual list of figures and items not yet covered.

Pages created

The implementation pass produced new pages in six areas:

  • People (58 new entity pages): an industry leadership tier (13), a politicians/regulators tier (11), a researcher tier (12), and a think-tank tier (20). This closed the gap previously described as naming researchers but not the executives and elected officials driving both industry and policy.
  • Companies (24 new): Big Tech and frontier labs (Microsoft, xAI, Mistral, Cohere, SSI, Thinking Machines, Perplexity, Hugging Face); Chinese AI (Zhipu, ByteDance/Doubao, Baidu/ERNIE); defense and specialized (Palantir, Anduril, Scale, Character.AI, Replika); and hardware (SMIC, ASML, SK Hynix, Samsung Semiconductor, Micron, AMD, Intel, Specialty Inference).
  • Models (15 new): Grok, Llama 3/4, Qwen 2.5, GLM-4, ERNIE 4, Mistral Large, the GPT-4 family, o1/o3, Gemini 2.5 Pro, Claude 3 Opus, Claude 3.5 Haiku, AlphaFold, Sora/Veo, and Doubao.
  • Concepts (33 new, plus one substantially expanded): major topics (mental health, deepfakes, copyright, risk taxonomy, AI alignment unifying); technical topics (reasoning/CoT, MoE, synthetic data, model collapse, prompt injection, jailbreaking/red teaming, C2PA/watermarking, multimodality, inference economics, context length); policy and societal topics (12 pages covering liability, sandbox, accountability, whistleblowing, elections, education, healthcare, journalism, surveillance, welfare, safety-vs-ethics, and compliance fragmentation); and infrastructure topics (data center siting, training data walls, CBRN uplift, jagged frontier, sovereign AI product, chip smuggling). AI Environmental Impact was expanded with PJM/ERCOT/Three Mile Island/Memphis/Louisiana detail.
  • Legislation (15 new): federal and state (TAKE IT DOWN Act, OMB M-24-10, OMB M-24-18, Utah SB 149, NYC LL 144, VA HB 2094, state deepfake statutes, CA SB 942/AB 3030/SB 896); and international (Canada AIDA, Brazil PL 2338, South Korea AI Basic Act, Japan AI Promotion Act, Australia voluntary).
  • Sources (9 new, plus 4 raw sources): Raine v. OpenAI, Garcia v. Character Technologies, APA AI chatbot guidance, the TAKE IT DOWN Act primary text plus reconstruction, and the Canada AIDA, Brazil PL 2338, South Korea AI Basic Act, Japan AI Promotion Act, and Australia voluntary text and analysis.

Residual gaps

People not yet covered

The bulk-coverage agents flagged figures referenced in newly created concept pages that do not yet have entity pages:

  • Timnit Gebru and Joy Buolamwini — referenced in AI Safety vs. AI Ethics Divide and Algorithmic Accountability and Bias Audits without entity pages.
  • David Chalmers and Eric Schwitzgebel — referenced in AI Welfare / Model Welfare / Moral Patienthood without entity pages.
  • Eliezer Yudkowsky and Nick Bostrom — foundational to existential-risk discourse and referenced widely without entity pages.
  • Roman Yampolskiy — periodically cited, with no page.
  • Tristan Harris and Aza Raskin — of the Center for Humane Technology, referenced obliquely.
  • Eric Schmidt — associated with SCSP and Schmidt Sciences and a principal funder, with no page.

These are candidates for the next gap pass.

Concept-page depth

Several new pages carry confidence: medium because they are broad surveys with thinner empirical grounding, and are candidates for source-driven reinforcement in a later ingest cycle:

Lint candidates

The bulk pass added cross-references to pages that do not yet exist (for example Gebru, Bostrom, Schmidt, and Llama 3 5 Haiku mentioned in some Llama context). A lint pass is needed to:

  • Convert orphan wikilinks to plain text or remove them.
  • Find missing back-references — for example, whether existing pages link to the new concept pages where appropriate.
  • Detect stub-level new pages and flag them for content expansion.
  • Audit sources_count accuracy across all 135 new pages.

Items from the original page not addressed

The implementation pass covered every page on the original list except two categories:

  • Stub-level company pages whose information density was judged too low to merit a full page: none — all listed companies received pages.
  • Items for which the bulk agents could not locate primary sources via WebFetch: the Australia mandatory-guardrails Proposals Paper text (referenced in the Australia voluntary AI Safety Standard page) and the EUROPA primary text for some EU implementing acts. Neither was treated as blocking.

Suggested next gap pass

After a lint cycle and 5–10 further source ingests, a fresh missing-coverage-2026-XX-XX.md could focus on:

  1. Content depth rather than page existence — which confidence: medium/low pages now have enough source support to upgrade to high.
  2. Cross-reference density — which structurally related pages are not linking to each other.
  3. Stale content — which last_updated timestamps are more than six months old without reinforcement.
  4. Specific people pages that the bulk pass left as residual (Gebru, Buolamwini, Yudkowsky, Bostrom, Chalmers, Schwitzgebel, Schmidt).

Relationships

Categories: Analysis & Trackers | meta | gap-analysis | coverage
What links here (2)