These visualizations are experimental and still being refined. Designs and data presentation may change.

Sector Attention by Risk Domain

Which industry sectors the governance documents addressing each risk domain actually name. Each cell is the percentage of the documents covering that domain which also mention that sector — a measure of attention breadth, not of how well the risk is handled.

Every row has its own denominator, so cells are comparable across a row but not down a column. A cell shows the share of the documents covering that risk domain which also mention the sector; the shading runs on a single 0–100% sequential scale. In the federal corpus the per-domain denominators are: 1. Discrimination & Toxicity 128 · 2. Privacy & Security 378 · 3. Misinformation 95 · 4. Malicious Actors & Misuse 256 · 5. Human-Computer Interaction 108 · 6. Socioeconomic & Environmental Harm 431 · 7. AI System Safety, Failures, & Limitations 390. A 50% cell in a row backed by 431 documents is far stronger evidence than a 50% cell in a row backed by 95, and column totals do not add up to anything meaningful because a document covering several domains is counted once in each of their rows. Switch to All Docs to widen the corpus from U.S. federal documents (2020–2026) to the full governance dataset. Click a cell to browse the governance documents for that domain and sector.

684 documents · 606 risk-classified

	Agri/Mining/Constr/Mfg Trade/Transport/Utilities Information Finance & Insurance Real Estate Prof & Technical Svcs Scientific R&D Mgmt/Admin/Support Educational Services Health Care & Social Arts/Entertainment Hospitality & Other Public Admin (excl NatSec) National Security
1 Discrimination & Toxicity (n=128)	15%	21%	57%	27%	19%	32%	48%	11%	41%	52%	13%	13%	66%	41%
2 Privacy & Security (n=378)	27%	26%	57%	19%	11%	35%	65%	7%	37%	31%	9%	7%	58%	71%
3 Misinformation (n=95)	8%	15%	71%	25%	14%	40%	35%	6%	37%	33%	17%	8%	68%	43%
4 Malicious Actors & Misuse (n=256)	27%	22%	63%	23%	9%	39%	60%	7%	33%	31%	11%	5%	52%	74%
5 Human-Computer Interaction (n=108)	16%	27%	56%	27%	19%	33%	58%	13%	50%	55%	22%	15%	57%	43%
6 Socioeconomic & Environmental Harm (n=431)	30%	24%	52%	16%	9%	34%	63%	7%	31%	24%	8%	6%	57%	71%
7 AI System Safety, Failures, & Limitations (n=390)	23%	21%	50%	16%	9%	32%	62%	7%	31%	32%	8%	6%	57%	63%

Each cell is the share of the documents covering that risk domain which also mention the sector. Each row is normalized to the documents covering that risk domain (its n is shown in the row label, and in full on hover) — compare across a row, not down a column. Click a cell to browse the governance documents for that domain and sector (browse spans the full governance dataset, not just the federal window).

Key Takeaways

1.National Security is the most-referenced sector overall, peaking at 74% of the documents covering Malicious Actors & Misuse.
2.Averaged across the seven domains, Public Administration excluding National Security leads at 59% while Management, Administrative, and Support Services trails at 8%.
3.Malicious Actors & Misuse is the most sector-concentrated domain, spanning 69 percentage points between its most- and least-referenced sectors.
4.Each row is normalized to its own denominator — 431 documents cover Socioeconomic & Environmental versus 95 for Misinformation, so compare across a row rather than down a column.

Sector and risk-domain labels come from the governance dataset’s own classification of each document; a document can name several sectors and several domains, and counts once in every pair it touches. Documents with no risk classification are excluded from every row.