Build on the AI-regulation dataset.

Ask a general-purpose chatbot which AI laws apply to you and it will produce a fluent, confident, unciteable answer — and sometimes a statute that does not exist. This API answers the same question from a curated corpus of primary sources, with no language model anywhere in the path. Every obligation, deadline and penalty we return is retrieved verbatim from a real record and carries the government URL it came from. Same input, same output, every time.

A live record from the deterministic engine — not a mock-up. This page renders it by calling the same code /v1/obligations runs, so it always matches what the endpoint would answer today.

{
  "law": "NYC Local Law 144 (Automated Employment Decision Tools)",
  "jurisdiction": "New York",
  "layer": "landscape",
  "lifecycle": "in_force",
  "in_force": true,
  "not_yet_effective": false,
  "effective_date": "2023-07-05",
  "applies_because": "Notable New York AI law, included for the national landscape. It binds businesses in New York — it may not apply to you unless you operate there.",
  "obligation": "Employers using automated employment decision tools to screen candidates or employees for NYC positions must conduct an annual independent bias audit, publish a summary on the employer website, notify candidates at least 10 business days in advance, and offer an alternative selection process on request.",
  "deadline": null,
  "penalty": "$500 for first violation; $500 to $1,500 for each subsequent violation, per day per candidate; DCWP enforcement",
  "status": null,
  "citation": "NYC Admin. Code Section 20-870 et seq.; 6 RCNY Sections 5-300 to 5-304 (effective July 5, 2023)",
  "government_body": "NYC Department of Consumer and Worker Protection",
  "may_not_apply": true,
  "source_url": "https://www.nyc.gov/site/dca/about/automated-employment-decision-tools.page",
  "source_tier": "primary"
}

Note source_url and source_tier: every obligation is traceable to the government page it came from, and you can tell a primary source from a secondary one. Informational only — not legal advice.

# List the 5 most recently updated records
curl "https://ai-law-tracker.com/api/v1/laws?limit=5&sort=updated_at&order=desc" \
  -H "X-API-Key: alt_free_9f3c…"

const res = await fetch(
  "https://ai-law-tracker.com/api/v1/laws?scope=state&jurisdiction=colorado",
  { headers: { "X-API-Key": process.env.ALT_API_KEY } }
);
const { data, meta } = await res.json();

1. Claude.ai → Settings → Connectors → Add custom connector.
2. URL: https://ai-law-tracker.com/api/mcp
3. Under Request headers, add x-api-key = your AI Law Tracker key (optional — skip it to try the anonymous tier).
4. Save, then ask e.g. “What AI hiring laws apply in California?”

1. Create a GPT → Configure → Actions → Import from URL.
2. Schema URL: https://ai-law-tracker.com/api/v1/openapi.json
3. Authentication → API Key, custom header name X-API-Key, paste your key.
4. In ChatGPT developer mode you can instead add the MCP endpoint above as an app.

X-API-Key: alt_free_9f3c…
# or
Authorization: Bearer alt_free_9f3c…

Two things to read before you quote a number from this dataset

1. Every matter is single_source. Each one is recorded from one official source and published with the URL it came from. verification is never withheld at any tier — read it, and follow source_url, before relying on any claim.

2. An empty result is a statement about us, not about the agency. Roughly half of the 254 registered sources are currently stale or disabled, and part of the registry is withheld pending review. /v1/enforcement/sources and /v1/enforcement/coverage both publish those numbers — including the withheld count — so coverage can be checked rather than assumed.

TIER-GATING: filters and sorting (scope, jurisdiction, status, in_force, updated_since, sort, order) are honoured at EVERY tier including anonymous — they are documented query semantics, so they are never silently dropped. What Free/anonymous callers get instead is low volume and shallow pagination: anon 5 rows per page with a trimmed field envelope, free-key 25 rows with the full field set. Developer+ raises the page size and unlocks deep pagination, the change feed and per-record history. When a request is capped, the response echoes meta.gated + an upgrade hint (it is never a hard error).

curl "https://ai-law-tracker.com/api/v1/laws?scope=state&jurisdiction=value" \
  -H "X-API-Key: alt_free_9f3c…"

{
  "api_version": "v1",
  "data": [
    {
      "id": "5f9c2b0a-1e34-4d21-9a77-0c1b2d3e4f56",
      "scope": "state",
      "jurisdiction": { "slug": "colorado", "name": "Colorado", "abbr": "CO" },
      "identifier": "SB 24-205",
      "title": "Consumer Protections for Artificial Intelligence",
      "record_type": "bill",
      "status": "Signed — effective 2026-02-01",
      "in_force": null,
      "record_date": "2024-05-17T00:00:00Z",
      "summary": "Requires developers and deployers of high-risk AI systems to use reasonable care to avoid algorithmic discrimination…",
      "source": "leg.colorado.gov",
      "official_url": "https://leg.colorado.gov/bills/sb24-205",
      "updated_at": "2026-07-05T14:02:11Z",
      "attribution": "Data by AI Law Tracker (CC BY 4.0) — https://ai-law-tracker.com"
    }
  ],
  "meta": { "total": 564, "count": 1, "limit": 20, "offset": 0 }
}

Every snapshot of the record (ascending observed_at), each with change_kind, changed_fields and the record values as of that snapshot. TIER-GATING: Developer+ only — anonymous/Free callers get 403 tier_upgrade_required. Lookback window climbs by tier: Developer 30 days, Pro 365 days, Business/Enterprise full history.

curl "https://ai-law-tracker.com/api/v1/laws/5f9c2b0a-…/history" \
  -H "X-API-Key: alt_free_9f3c…"

Time-ordered feed of every observed insert/update across the dataset, unioned from the append-only history of legal records and jurisdiction summaries. Poll with ?since=<meta.cursor>. TIER-GATING: Developer+ only — anonymous/Free callers get 403 tier_upgrade_required (the change feed is the paid differentiator). The lookback window is clamped per tier (Developer 30-day, Pro 90-day, Business/Enterprise full) — see meta.window.

curl "https://ai-law-tracker.com/api/v1/changes?since=value&scope=state" \
  -H "X-API-Key: alt_free_9f3c…"

{
  "api_version": "v1",
  "data": [
    {
      "event_id": "a1b2c3d4-0000-4444-8888-abcdef012345",
      "entity": "law_record",
      "record_id": "5f9c2b0a-1e34-4d21-9a77-0c1b2d3e4f56",
      "scope": "state",
      "jurisdiction": { "slug": "washington", "name": "Washington", "abbr": "WA" },
      "identifier": "HB 1170",
      "change_kind": "update",
      "changed_fields": ["status"],
      "field_changes": [{ "field": "status", "from": "In committee", "to": "Passed House" }],
      "observed_at": "2026-07-07T09:15:44.512Z",
      "attribution": "Data by AI Law Tracker (CC BY 4.0) — https://ai-law-tracker.com"
    }
  ],
  "meta": {
    "count": 1, "limit": 50, "offset": 0, "order": "desc",
    "cursor": "2026-07-07T09:15:44.512Z",
    "window": { "tier": "free", "lookback_days": 90, "clamped": false }
  }
}

Convenience view over /changes with significant=true and no value diffs — the newest meaningful changes, human-scannable. TIER-GATING: same as /changes — Developer+ only (403 tier_upgrade_required for anonymous/Free).

curl "https://ai-law-tracker.com/api/v1/feed?since=value&scope=state" \
  -H "X-API-Key: alt_free_9f3c…"

The audited AI-regulation news feed (headline + short excerpt + publisher + source URL — never full article text). Free/anon get a TASTE: fewer results, shallow pagination, only the newest headlines (a recency window) and the excerpt withheld. Developer+ gets the full archive, high volume and the excerpt. meta.window shows the tier recency floor; meta.gated + meta.upgrade show what was limited.

curl "https://ai-law-tracker.com/api/v1/news?scope=state&jurisdiction=value" \
  -H "X-API-Key: alt_free_9f3c…"

The interpreted "what must I do" layer — who must do what, by when, or owe what penalty — for the given context. Every obligation is grounded in a real law record + its primary source (no LLM). Provide at least a jurisdiction or a sector. NOT legal advice. Paid feature (Pro+); lower tiers receive a one-item preview (meta.preview=true) with an upgrade hint.

curl "https://ai-law-tracker.com/api/v1/obligations?jurisdiction=value&sector=value" \
  -H "X-API-Key: alt_free_9f3c…"

Per applicable law: the monetary/other penalty, enforcing body, citation, and primary source, plus a sector-level penalty-structure summary (full result only). Provide at least a jurisdiction or a sector. Paid feature (Developer+); lower tiers receive a one-item preview.

curl "https://ai-law-tracker.com/api/v1/penalties?jurisdiction=value&sector=value" \
  -H "X-API-Key: alt_free_9f3c…"

The compliance-report engine (deterministic, no LLM — cannot hallucinate) as an API. POST a business profile; get the applicable AI-law obligations plus a reproducible 1–10 risk score with a named factor breakdown. state + sector are required. Paid feature (Pro+); lower tiers receive a preview (risk band + counts + one sample obligation; factor breakdown withheld).

curl "https://ai-law-tracker.com/api/v1/assess" \
  -X POST \
  -H "X-API-Key: alt_free_9f3c…"

Submit an email to mint a free-tier key. The plaintext key is emailed to that address (your durable copy — find it again there anytime) AND returned once in this response. We never store the plaintext server-side. One active free key per email; abusive request rates are throttled per IP.

curl "https://ai-law-tracker.com/api/v1/keys" \
  -X POST -H "Content-Type: application/json" \
  -d '{"email":"you@company.com"}'

{
  "api_version": "v1",
  "key": "alt_free_9f3c…  (shown once on screen; also emailed to you)",
  "key_prefix": "alt_free_9f3c",
  "id": "7c0d…",
  "tier": "free",
  "email": "you@company.com",
  "email_delivery": { "delivered": true },
  "note": "Copy your key now — shown once on this screen. We also emailed it to you, so you can always find it again in that email."
}

Returns the presenting key's own account — tier, effective rate/quota limits, live durable usage (requests used today + this month, from the api_usage counter), remaining, reset times, and the capabilities the tier unlocks. The key IS the credential; no separate login. Powers the self-serve dashboard at /developers/dashboard. This call itself counts as one request.

curl "https://ai-law-tracker.com/api/v1/me" \
  -H "X-API-Key: alt_free_9f3c…"

{
  "api_version": "v1",
  "data": {
    "key_prefix": "alt_free_9f3c",
    "tier": "free",
    "owner": "y••••@company.com",
    "revoked": false,
    "created_at": "2026-07-08T12:00:00Z",
    "last_used_at": "2026-07-09T14:22:07Z",
    "limits": { "per_min": 60, "per_day": 2000, "per_month": 60000 },
    "usage": {
      "tracking": true,
      "day":   { "used": 137, "limit": 2000,  "remaining": 1863,  "resets_in_seconds": 35080 },
      "month": { "used": 4210, "limit": 60000, "remaining": 55790, "resets_in_seconds": 2110480 }
    },
    "capabilities": {
      "list_max": 25, "full_fields": true, "advanced_filters": true,
      "changes_window_days": null, "history_window_days": null,
      "webhooks": 0, "bulk_export": false
    }
  }
}

Self-service revoke of the presenting key. Irreversible — the key can no longer authenticate afterwards. Create a new key to keep using the API.

curl "https://ai-law-tracker.com/api/v1/me" \
  -X DELETE \
  -H "X-API-Key: alt_free_9f3c…"

{
  "api_version": "v1",
  "data": {
    "key_prefix": "alt_free_9f3c",
    "tier": "free",
    "owner": "y••••@company.com",
    "revoked": false,
    "created_at": "2026-07-08T12:00:00Z",
    "last_used_at": "2026-07-09T14:22:07Z",
    "limits": { "per_min": 60, "per_day": 2000, "per_month": 60000 },
    "usage": {
      "tracking": true,
      "day":   { "used": 137, "limit": 2000,  "remaining": 1863,  "resets_in_seconds": 35080 },
      "month": { "used": 4210, "limit": 60000, "remaining": 55790, "resets_in_seconds": 2110480 }
    },
    "capabilities": {
      "list_max": 25, "full_fields": true, "advanced_filters": true,
      "changes_window_days": null, "history_window_days": null,
      "webhooks": 0, "bulk_export": false
    }
  }
}

Create a webhook. When a matching legal record change is observed, an HMAC-SHA256-signed POST is delivered to target_url. The signing_secret is returned ONCE in this response. Verify each delivery (see x-webhooks). TIER-GATING: Pro+ only (Free/Developer allowance = 0 → 403 tier_upgrade_required). Per-key subscription cap: Pro 3, Business 25, Enterprise unlimited.

curl "https://ai-law-tracker.com/api/v1/webhooks" \
  -X POST \
  -H "X-API-Key: alt_pro_…" \
  -H "Content-Type: application/json" \
  -d '{"target_url":"https://your-app.com/hooks/alt"}'

{
  "api_version": "v1",
  "data": {
    "id": "b3c4…",
    "target_url": "https://your-app.com/hooks/alt",
    "events": ["law.created", "law.updated"],
    "active": true,
    "signing_secret": "whsec_… (shown once)",
    "created_at": "2026-07-08T12:00:00Z"
  }
}

The machine-readable form of the public /accuracy page. Returns the latest source-liveness sweep summary (how many cited sources were checked, verified live, confirmed broken, or unverified), the count of confirmed-broken sources in the review queue, and the total append-only change-observations on record. Free/public: no tier gate, rate-limited like any endpoint. Every figure is a live count over real rows.

curl "https://ai-law-tracker.com/api/v1/accuracy" \
  -H "X-API-Key: alt_free_9f3c…"

AI-related regulatory enforcement (FTC/DOJ/SEC/state AG actions, etc.) and civil litigation (federal court dockets), over the same six AET fields — agency, respondent, allegation, outcome, penalty, statutory basis — plus forum for litigation rows. `change_type` tells the two lanes apart. ONLY status approved/published rows are ever returned: every draft the automated watch crons write starts pending and is invisible here until reviewed, so an empty data[] with meta.total: 0 is the expected response while a batch awaits approval — not an error. TIER-GATING: Enterprise-only. Anonymous, Free, and every self-serve paid tier (Developer $29 / Pro $99 / Business $299) get a hard 403 tier_upgrade_required — there is no teaser and no partial access below Enterprise. Same hard-gate pattern as /v1/changes, /v1/feed, and /v1/webhooks.

curl "https://ai-law-tracker.com/api/v1/enforcement?scope=state&jurisdiction=value" \
  -H "X-API-Key: alt_free_9f3c…"

The free awareness OVERVIEW of a sector (name, risk, enforcement bodies, federal penalty structure) plus the DERIVED per-law obligations & penalties. The overview is open to every tier; the derived obligations/penalties are the interpreted layer — full for Pro+, a one-item preview (meta.preview=true) + upsell below. NOT legal advice.

curl "https://ai-law-tracker.com/api/v1/sectors/{sector}" \
  -H "X-API-Key: alt_free_9f3c…"

A forward-looking compliance calendar derived deterministically from the curated corpus — each entry carries the VERBATIM source deadline text + primary-source URL (no fabricated dates; a jurisdiction whose deadline text has no explicit date is only counted in meta.undated). Free for every tier. Pass ?format=ical (alias ics) for an RFC 5545 .ics you can subscribe to (webcal). The interpreted "what must I do about each deadline" layer is on /obligations + /assess.

curl "https://ai-law-tracker.com/api/v1/deadlines?scope=state&jurisdiction=value" \
  -H "X-API-Key: alt_free_9f3c…"

The named, name-searchable AI laws (valid slugs for /bills/{slug}), from the primary-sourced landmark registry. Free.

curl "https://ai-law-tracker.com/api/v1/bills" \
  -H "X-API-Key: alt_free_9f3c…"

One curated bill — core fields, tiered source URLs, citations (identifier + audited alternate names), and cross-links. Where a matching legal_records row exists, the response binds the live change timeline (links.history → /laws/{id}/history). Primary-sourced; the DB binding is best-effort (a miss omits the link, never fabricates it).

curl "https://ai-law-tracker.com/api/v1/bills/{slug}" \
  -H "X-API-Key: alt_free_9f3c…"

A record's own official/.gov link plus the curated jurisdiction sources, each tiered primary/secondary. Basic tiers (anon/Free) get primary sources only; Developer+ get the full set incl. secondary/analysis links. Only real held URLs — never fabricated.

curl "https://ai-law-tracker.com/api/v1/laws/5f9c2b0a-…/sources" \
  -H "X-API-Key: alt_free_9f3c…"

A record's identifier, title, audited alternate names, and a formatting-convenience suggested citation over public metadata (labelled — never an official cite). Free at every tier.

curl "https://ai-law-tracker.com/api/v1/laws/5f9c2b0a-…/citations" \
  -H "X-API-Key: alt_free_9f3c…"

A conventional token endpoint over the same key store as POST /keys. Submit an email; get a free-tier token, emailed once (durable copy) and returned once here. Only ever mints a FREE token (no privilege escalation); the owner/HMAC secret is never exposed. Paid tiers are granted via Polar checkout (POST /checkout). Throttled per IP; one active token per email.

curl "https://ai-law-tracker.com/api/v1/auth/token" \
  -X POST \
  -H "X-API-Key: alt_free_9f3c…"

Generate the personalized, source-grounded AI-law compliance report for a business profile — the SAME deterministic engine behind the $79 /compliance-report product (no LLM in the grounded path — it cannot invent a statute, deadline, or penalty). POST state + sector (+ optional AI-use profile). Default response is JSON: the risk assessment (1–10 score + obligations) plus a signed `report.download_url` for the rendered PDF. Pass `format:"pdf"` (or send `Accept: application/pdf`) to stream the PDF binary inline instead. Paid feature (Pro+) with a monthly report quota per tier (Pro 5/mo, Business 25/mo); when the quota is exhausted the endpoint returns 402 with an upgrade / single-report hint. Lower tiers receive a preview (risk band + counts + one sample obligation) with the PDF/download withheld behind an upgrade hint. Informational only — NOT legal advice.

curl "https://ai-law-tracker.com/api/v1/report" \
  -X POST \
  -H "X-API-Key: alt_free_9f3c…"

Statutes, regulations and federal executive instruments at title/chapter/part level, each with its official citation, status, effective date, the official government source URL, and a many-to-many classification into 33 branches and 227 subcategories. Records LINK to the official text; they do not reproduce it. METERED ON RECORDS SERVED, not just on requests. A request cap stops a fast client; a record cap stops a patient one, and only the second is priced against the size of the corpus. Every response carries `meta.record_budget` with what is left in the day and 30-day windows; exhausting it returns 429 `record_budget_exceeded` (distinct from 429 `rate_limited`). PRODUCT SCOPE: this endpoint belongs to the US Law Platform. A key licensed only for AI Law Tracker (`product: ai_law`) is NOT refused here — it is served at the PUBLIC (keyless) capability level, the same few records any anonymous caller gets. The response headers say so explicitly: `X-ALT-Product` (what your key holds), `X-ALT-Product-Surface` (`us_law`) and `X-ALT-Product-Scope` (`in_product` | `out_of_product`). Your request rate and daily/monthly quota are NEVER degraded — those are metered per key wherever you call. What degrades is capability: page size, pagination depth, field envelope and the record budget.

curl "https://ai-law-tracker.com/api/v1/us-law?branch=value&branch_any=value" \
  -H "X-API-Key: alt_free_9f3c…"

Websearch-syntax full text ("quoted phrases", -exclusion) with a citation-substring fallback; the strategy that answered is reported in `meta.strategy`. A query that is a presidential document number ("EO 14110", "Proclamation 10925") is resolved against the INDEXED number column only — a miss returns nothing with an explanatory `meta.notice`, never a guessed match, because misidentifying a presidential document is worse than returning nothing. METERED ON RECORDS SERVED, not just on requests. A request cap stops a fast client; a record cap stops a patient one, and only the second is priced against the size of the corpus. Every response carries `meta.record_budget` with what is left in the day and 30-day windows; exhausting it returns 429 `record_budget_exceeded` (distinct from 429 `rate_limited`). PRODUCT SCOPE: this endpoint belongs to the US Law Platform. A key licensed only for AI Law Tracker (`product: ai_law`) is NOT refused here — it is served at the PUBLIC (keyless) capability level, the same few records any anonymous caller gets. The response headers say so explicitly: `X-ALT-Product` (what your key holds), `X-ALT-Product-Surface` (`us_law`) and `X-ALT-Product-Scope` (`in_product` | `out_of_product`). Your request rate and daily/monthly quota are NEVER degraded — those are metered per key wherever you call. What degrades is capability: page size, pagination depth, field envelope and the record budget.

curl "https://ai-law-tracker.com/api/v1/us-law/search?q=value&branch=value" \
  -H "X-API-Key: alt_free_9f3c…"

The 33 branches and 227 subcategories the corpus is classified into, with descriptions, so a consumer can build ?branch= / ?domain= filters without guessing. Deliberately OPEN to everyone and NOT metered: it DESCRIBES the corpus, it is not the corpus, and a classification a customer cannot inspect is a black box. Also returns `meta.health` — `records_total`, `records_placeholder` (records still in the `other_unclassified` bucket) and `avg_domains_per_record`. That placeholder count is published rather than hidden: it is the honest measure of how far the classification actually reaches, and it is the number every public coverage claim on the site is checked against.

curl "https://ai-law-tracker.com/api/v1/us-law/taxonomy" \
  -H "X-API-Key: alt_free_9f3c…"

One row per jurisdiction: record counts by instrument type, the code edition held, and the verification stamps. Aggregated in the database — this endpoint must never be answered by paging the corpus, since inviting a caller to walk 31,800 records just to count them is exactly the bulk copy the record budget exists to prevent. Open for the same reason as /us-law/taxonomy: coverage is how you evaluate the dataset before paying for it.

curl "https://ai-law-tracker.com/api/v1/us-law/jurisdictions?jurisdiction_level=federal" \
  -H "X-API-Key: alt_free_9f3c…"

Published opinions filed 2015-01-01 onward from federal appellate courts and state supreme/appellate courts. Each record IDENTIFIES a decision — case name, court, filing date, docket number, reporter citation, and how many later opinions cite it — and links to the full decision at `absolute_url`. THERE IS NO OPINION TEXT AND NO HOLDING IN THIS CORPUS. `q` is a websearch full-text match over the CASE NAME only (quoted phrases and -exclusion work: `"Trump v. Cook"`, `monsanto -durnell`), and every response that ran one says so in `meta.search_field`. Do not describe this endpoint as searching the text of decisions. METERED ON RECORDS SERVED, on the SAME budget as /v1/us-law rather than a second one. That is deliberate: a separate allowance would have widened every existing key the day this dataset shipped. Every response carries `meta.record_budget`; exhausting it returns 429 `record_budget_exceeded`. TWO LIMITS COME FROM THE SIZE OF THE TABLE, NOT FROM YOUR TIER, and they apply identically at every tier including Enterprise. (1) `meta.total` is the PLANNER’S ESTIMATE above 1,000 rows — it measured 53% low on one full-text query — so it is always accompanied by `meta.total_is_estimate`; pass `count=exact` for the true number, at up to ~3.3 s. (2) `sort=citation_count` and `cited_min` have no index behind them and are refused with 400 unless the query already carries a narrowing filter; unfiltered, they are cancelled by the database statement timeout. PRODUCT SCOPE: this endpoint belongs to the US Law Platform. A key licensed only for AI Law Tracker is served here at the public (keyless) level, reported in `X-ALT-Product-Scope` — not refused.

curl "https://ai-law-tracker.com/api/v1/us-law/opinions?court=value&court_level=value" \
  -H "X-API-Key: alt_free_9f3c…"

The complete record for one CourtListener cluster id — the `cluster_id` field of any list result. Provenance (absolute_url, source_authority, bulk_edition, retrieved_at) is never withheld at any tier: the link to the decision is how you check our work, and withholding it to build an upsell would mean selling an unverifiable claim. The record identifies the decision; it does not contain it. `meta.corpus.full_text_at` points at the opinion itself. To fetch many at once use /us-law/opinions?cluster_id=1,2,3 (up to 100).

curl "https://ai-law-tracker.com/api/v1/us-law/opinions/{cluster_id}" \
  -H "X-API-Key: alt_free_9f3c…"

One row per REGISTERED court: how many opinions we hold, how many carry a reporter citation, and the earliest and latest filing dates. It is both the coverage table and the lookup the `court=` filter needs — a filter whose valid values are undiscoverable is not a filter, which is why this is open. 293 courts are registered and 146 hold opinions, so 147 rows come back with `opinions: 0`. They are not hidden: omitting them would overstate coverage. Read the date spans too — scotus shows 37,792 rows because CourtListener publishes orders and summary dispositions as clusters, and those thin out sharply after 2020. THE COUNTS ARE MATERIALISED AND THE RESPONSE SAYS WHEN. Computed live the aggregate takes 2.8 s against 988,554 rows, which on a keyless endpoint would be a self-inflicted denial-of-service. So every response carries `meta.counts_refreshed_at`, and `meta.counts_stale` goes true once the rollup is more than 48 hours old (the refresh runs daily). The numbers are never presented as current — they are presented with the timestamp that says how current they are.

curl "https://ai-law-tracker.com/api/v1/us-law/courts?court_level=federal_appellate&jurisdiction_level=federal" \
  -H "X-API-Key: alt_free_9f3c…"

Plans a bulk export of any slice /us-law can filter and returns a MANIFEST: the exact row count, what it will cost against your record allowance, and one SIGNED, short-lived download URL per part (NDJSON | CSV | JSON). The bytes come from the signed part URLs; this call serves no records and charges nothing. NOT SOLD SELF-SERVE AT ANY PRICE. Taking the whole corpus as a file is a licensing conversation, not a checkout — every self-serve plan keeps full API access to the same data and pages it instead. A below-tier caller gets 403 `tier_upgrade_required` naming the tier it opens at. Part URLs carry a signed token rather than your API key, so a link can be handed to a loader without handing over your credential; every download is still metered against the key that created it. PRODUCT SCOPE: this endpoint belongs to the US Law Platform. A key licensed only for AI Law Tracker (`product: ai_law`) is NOT refused here — it is served at the PUBLIC (keyless) capability level, the same few records any anonymous caller gets. The response headers say so explicitly: `X-ALT-Product` (what your key holds), `X-ALT-Product-Surface` (`us_law`) and `X-ALT-Product-Scope` (`in_product` | `out_of_product`). Your request rate and daily/monthly quota are NEVER degraded — those are metered per key wherever you call. What degrades is capability: page size, pagination depth, field envelope and the record budget.

curl "https://ai-law-tracker.com/api/v1/us-law/export?format=ndjson&fields=full" \
  -H "X-API-Key: alt_free_9f3c…"

The complete record for one record_uid, including provenance (official_url, source_authority, full_text_url, full_text_sha256, retrieved_at, last_verified_at). Provenance is never withheld at any tier: on a legal surface the citation and the link to the government source are how you check our work, and withholding them to create an upsell would mean selling an unverifiable claim. To fetch many at once use /us-law?record_uid=uid1,uid2 (up to 100).

curl "https://ai-law-tracker.com/api/v1/us-law/records/{record_uid}" \
  -H "X-API-Key: alt_free_9f3c…"

The recorded change log for one record: one entry per changed FIELD, newest first, across status, dates, citation, identifier, title, instrument type, classification, source URLs and summary. READ meta.capture_started_at BEFORE READING data. Nothing recorded a record-level diff on this corpus before 2026-08-20, and there is deliberately NO BACKFILL, because a backfill would have to invent what the record used to say. An empty result therefore means "we did not observe a change since capture began" and says so in meta.notice — it is NEVER a claim that the law has not changed. TIER: sampled below Pro (the most recent few entries, with the true total stated), full at Pro and above on either product ladder.

curl "https://ai-law-tracker.com/api/v1/us-law/records/{record_uid}/history" \
  -H "X-API-Key: alt_free_9f3c…"

One row per jurisdiction holding law in a branch or subcategory, with record counts, how many are in force, the split by instrument type, the latest effective date and when we last verified it. This is the answer you would otherwise have to page the entire 31,867-record corpus to assemble. COUNTING. `records` counts EVERY domain assignment, primary or secondary; `records_primary` counts only records whose PRIMARY classification is this branch. Both are reported because they differ a lot (health_law: 2,852 vs 1,942) and a comparison built on primary assignments alone would report that a state has no privacy law when its privacy statutes happen to be filed under consumer protection first. TIER. Ordinary single-field filtering on /us-law is open at every tier including keyless — discovery is never the thing that is metered. This is the assembled answer, so below Pro it returns a real 5-jurisdiction SAMPLE with the true totals stated in meta, and the full table at Pro and above on either product ladder.

curl "https://ai-law-tracker.com/api/v1/us-law/compare?branch=value&domain=value" \
  -H "X-API-Key: alt_free_9f3c…"

Streams a single part of an export planned by GET /us-law/export. Authenticated by the SIGNED TOKEN in the manifest, not by an API key — so a part URL can be handed to a loader without handing over your credential. Tokens are short-lived and each download is still metered against the key that created the manifest. Any API key sent alongside is ignored: authority here comes from the token.

curl "https://ai-law-tracker.com/api/v1/us-law/export/download?token=value" \
  -H "X-API-Key: alt_free_9f3c…"

AI-related risk-factor disclosures extracted from SEC EDGAR filings. ENTERPRISE-ONLY: there is no teaser and no partial access on any self-serve tier — a below-tier caller gets 403 tier_upgrade_required, never a trimmed response. Documented here so the capability is discoverable; the gate is in the route, not in this spec.

curl "https://ai-law-tracker.com/api/v1/disclosures?q=value&company=value" \
  -H "X-API-Key: alt_free_9f3c…"

Regulatory enforcement actions and private litigation involving AI, at MATTER level: forum, agency, parties, status, relief and the statutes invoked. One row per dispute, not per press release — an agency action with a complaint, a settlement and a final order is three rows on /v1/enforcement and ONE matter here. OPEN AT EVERY TIER since 2026-08-22. It used to answer a hard 403 below Enterprise, which made the dataset invisible and therefore unsellable. What is metered instead is DEPTH: the assembled chain on /enforcement/matters/{matter_uid} — parties, relief amounts, the event timeline and the cited statutes — is sampled below the paid step. A record budget would have been theatre on a 91-row registry. A SECOND GATE THAT IS NOT ABOUT MONEY: 30 of 91 matters are duplicates of another row or still carry an untranslated CAPTION, and are withheld from every caller not licensed for the raw registry — `meta.registry.withheld_not_publish_ready` reports how many THIS query hid. A further 6 ARE published with one descriptive field nulled and named in `unavailable_fields`; withholding a $1.4bn settlement because its forum_name has an untranslated clause would have been a worse answer. See /enforcement/coverage for the full breakdown. READ `verification` BEFORE QUOTING ANYTHING. Every matter currently held is `single_source`. An empty data[] means no matter matched — never that an agency has taken no action. /enforcement/sources says which of the 254 monitored sources are actually answering.

curl "https://ai-law-tracker.com/api/v1/enforcement/matters?jurisdiction=value&jurisdiction_level=value" \
  -H "X-API-Key: alt_free_9f3c…"

The matter plus its EVENTS (timeline), PARTIES, RELIEF, statutory CITATIONS and the SOURCE DOCUMENTS each claim traces to. This is the dossier the product is sold on. THE CHAIN IS WHAT IS METERED. Below the paid step each chain is a 2-row sample and `data.chain_totals` reports the true size of every one of them — never an empty wall, and never a sample you cannot tell is a sample. Listing and searching matters stays fully open. FIELDS THAT LOOK LIKE HEDGES AND ARE THE PRODUCT: `events[].is_date_estimated` (the source gave a year, not a day), `verification`, `laws[].link_status` (`unlinked` = we hold the citation verbatim but have not resolved it — a confidently wrong citation is worse), and `duplicate_of`. PARTIES ARE REDACTED BY POLICY, in the database, not here: 97 of 235 party rows publish and the rest are withheld — named private individuals and minors by default. See the pii_class and redaction_policy vocabularies on /enforcement/categories. A 404 is returned both for a matter that does not exist AND for one withheld pending review, on purpose: distinguishable codes would turn the readiness gate into an enumeration oracle.

curl "https://ai-law-tracker.com/api/v1/enforcement/matters/{matter_uid}" \
  -H "X-API-Key: alt_free_9f3c…"

The 254 official sources behind the enforcement layer — agency newsrooms, AG press feeds, court dockets — with the last successful read, the fault state and whether each is stale against its own silence budget. IT PUBLISHES THE BAD NEWS. 132 of 254 are enabled and 123 are stale (measured 2026-08-22). Those go out as they are, on the same argument /v1/us-law/jurisdictions publishes its OLDEST verification stamp: an enforcement product whose coverage table only showed the healthy sources would be advertising, not data. FAULTS ARE CLASSIFIED, NOT QUOTED. `last_fault_reason_code` is a closed vocabulary (blocked, unauthorized, not_found, timeout, adapter_missing, egress_unavailable, other) with a one-line label; the raw internal fault text never leaves the server. Read this before concluding anything from an empty /enforcement/matters result: a stale source is a gap in OUR monitoring, never a statement that an authority has taken no action.

curl "https://ai-law-tracker.com/api/v1/enforcement/sources?jurisdiction_level=federal&is_enabled=value" \
  -H "X-API-Key: alt_free_9f3c…"

Thirteen closed vocabularies, 104 terms: matter_type, matter_status, forum_type, party_role, party_kind, relief_type, event_type, ai_role, law_relation, verification, jurisdiction_level, pii_class and redaction_policy. `code` is the API value; `label` and `description` are for display. These are the exact values the /enforcement/matters filters accept, which is why the endpoint is open: a documented filter whose valid values are undiscoverable is not a filter. TWO OF THEM DESCRIBE WHAT WE DO NOT PUBLISH. pii_class and redaction_policy are the terms behind the party gate that withholds 138 of 235 party rows. Publishing the vocabulary is how that policy becomes checkable rather than merely asserted.

curl "https://ai-law-tracker.com/api/v1/enforcement/categories?kind=value" \
  -H "X-API-Key: alt_free_9f3c…"

Counts the registry rather than serving it: matters held, matters publishable, and how many are withheld as duplicates or for untranslated internal language — plus breakdowns by matter type, jurisdiction level and agency, and the de-duplicated monetary relief total. THE WITHHELD COUNTS ARE THE POINT. A caller who sees `total: 61` on /enforcement/matters would reasonably conclude the registry IS 61 matters. It is 91: 11 duplicates and 21 untranslated captions (overlapping) leave 30 withheld, and 6 of the published ones carry a nulled field. This is the same commitment /v1/us-law/taxonomy makes by publishing `records_placeholder`. The relief total is computed over PUBLISHABLE, de-duplicated matters only and says so in `relief.basis`. Summed naively across all 91 rows it is wrong by at least the $3.2M DOJ/OpenAI settlement, which is present twice under two matter_uids.

curl "https://ai-law-tracker.com/api/v1/enforcement/coverage" \
  -H "X-API-Key: alt_free_9f3c…"

const crypto = require('crypto');
function verify(req, secret) {
  const ts = req.headers['x-alt-timestamp'];
  const sig = (req.headers['x-alt-signature'] || '').replace(/^sha256=/, '');
  const expected = crypto.createHmac('sha256', secret).update(`${ts}.${req.rawBody}`).digest('hex');
  const a = Buffer.from(sig), b = Buffer.from(expected);
  return a.length === b.length && crypto.timingSafeEqual(a, b);
}

{
  "api_version": "v1",
  "error": {
    "code": "rate_limited",
    "message": "Rate limit exceeded for the minute window. Retry in 42s.",
    "docs": "https://ai-law-tracker.com/developers",
    "retry_after": 42
  }
}