Data feed · Compliance & enforcement

State attorney general enforcement actions feed

50 AG offices · daily sweep · from $2,000 / mo

Coverage
50 states + DC
Cadence
Daily
Typical volume
~2k records
Difficulty
High
Price
$2,000 / mo

What's in it

Enforcement activity from all 50 state attorney general offices, swept daily and turned from prose into records:

Every extracted field carries the source sentence it came from, and anything the announcement does not state is null rather than inferred. Low-confidence extractions route to human review before they publish rather than shipping with a confidence score and a shrug.

Delivery

Daily sweep, delivered as new and updated actions. Lands in S3, a webhook, or your database. The manifest names which AG offices responded on each run, because an office that published nothing and an office whose site changed look the same until you measure them separately.

Why this one is hard

This is the cheapest source in the compliance stack to fetch and the most expensive to make useful, and the gap between those two facts is the entire product. The acquisition is 50 press release sections, several of which expose RSS. Anyone can do that part in an afternoon.

The output is prose written by a communications office. An announcement says that the attorney general reached a settlement with several named companies over conduct described in a paragraph, for an amount mentioned once, possibly in words rather than digits. Turning that into a defendant list, a category, and a number is the work, and it does not yield to regular expressions alone.

Extraction that guesses is worse than no extraction. A fabricated defendant name in a screening feed is the worst output this kind of system can produce, because it is confident, plausible, and lands in a queue where somebody acts on it. The rule here is grounding: the model sees only the article text, every extracted value carries the sentence that supports it, anything unsupported comes back null, and below a confidence threshold a human looks at it before it ships. That is slower and it is the only defensible way to do it.

AG sites redesign constantly and without notice, more often than any government source in the catalogue. They are communications properties, not records systems, so they follow web fashion. Expect to re-point selectors several times a year and budget for it rather than being surprised.

Archives are shallow. Many offices paginate ten pages deep with no date filter and no stable permalink structure, which makes the initial backfill a fundamentally different job from the daily delta. A few offices publish nothing structured at all, and those are declared as partial coverage rather than quietly dropped.

Build versus buy

Build it yourself if you care about two or three states and you are willing to read the announcements. At that scale a human skimming an RSS feed genuinely outperforms a pipeline.

Buy it when you need all 50 on a daily cycle keyed to an entity list, because the recurring cost is not the crawl. It is the extraction quality bar, the review queue, and re-pointing selectors after each redesign. That is a standing commitment of a few hours a month that never goes away, against $2,000/mo.

One use restriction: enforcement data delivered here is not a consumer report and is not licensed for FCRA-covered decisions about employment, credit, insurance, or housing.

Related

This feed is the enforcement layer of the Compliance Screening Stack ($3,650/mo), sitting alongside OFAC sanctions, FINRA BrokerCheck, and Secretary of State entity status, all resolved to the same party model.

Want to know what extraction quality actually looks like on this source? Ask for a sample and I'll come back inside 48 hours with a free Feed Feasibility Check: what's acquirable, rough cost per record, and my honest read on build versus buy. - Andrew
Get a quote for this feed → See the Compliance Stack →