id stringlengths 5 5 | framework stringlengths 4 56 | type stringlengths 6 37 | about stringlengths 52 164 | comes_up_in stringlengths 5 63 |
|---|---|---|---|---|
fw-01 | HIPAA | US federal law | Protects health information held by covered entities and their business associates, and sets two ways to de-identify it: Safe Harbor and Expert Determination. | Healthcare administration, insurance, biotech and labs |
fw-02 | 42 CFR Part 2 | US federal rule | Adds stricter protection to substance use disorder treatment records. | Healthcare administration |
fw-03 | CCPA and CPRA | US state law (California) | Gives California residents rights over personal information, including employee, candidate and business-contact data, and sets duties for businesses that handle it. | Almost every industry |
fw-04 | Other US state privacy laws | US state laws | A growing set of state laws with similar rights and duties; which apply depends on where people live and your size. | Almost every industry |
fw-05 | GDPR | EU regulation | Sets rules for personal data of people in the European Union, including lawful basis, minimisation and what counts as anonymous. | Any industry with EU customers, staff or contacts |
fw-06 | UK GDPR | UK law | The United Kingdom's version of the GDPR, with its own regulator. | Any industry with UK customers, staff or contacts |
fw-07 | PIPEDA and Quebec Law 25 | Canadian federal and provincial law | Canada's private-sector privacy rules, and Quebec's stricter law with specific rules on de-identification and anonymization. | Any industry with Canadian customers or staff |
fw-08 | GLBA and the FTC Safeguards Rule | US federal law and rule | Requires financial institutions, a group that includes some accounting and tax firms, to protect customer financial information. | Finance, insurance, accounting |
fw-09 | FCRA | US federal law | Governs consumer reports, such as credit, background and tenant screening reports, and limits how they are used and shared. | Staffing and recruiting, property management, finance |
fw-10 | COPPA | US federal law | Protects personal information of children under 13 online and requires parental consent in many cases. | Gaming, software, retail and e-commerce |
fw-11 | PCI DSS | industry standard | A security standard for payment card data set by the card networks; a contract requirement, not a law. | Retail and e-commerce, finance |
fw-12 | CAN-SPAM | US federal law | Sets rules for commercial email, including opt-outs. | Agencies, retail and e-commerce |
fw-13 | Fair Housing Act | US federal law | Prohibits housing discrimination on protected characteristics, which matters when notes mention them. | Property management |
fw-14 | Driver's Privacy Protection Act | US federal law | Limits use and disclosure of personal information from motor vehicle records. | Logistics and freight |
fw-15 | ITAR and EAR | US export control regimes | Restrict who may receive certain technical data and items, including by sharing a file. | Aerospace, semiconductors, manufacturing |
fw-16 | CUI and DFARS | US government requirements | Rules for handling controlled unclassified information on defense contracts. | Aerospace, manufacturing |
fw-17 | GINA | US federal law | Limits use of genetic information in health insurance and employment. | Biotech and labs, insurance |
fw-18 | 21 CFR Part 11 | US federal rule | Sets requirements for electronic records and signatures in FDA-regulated work. | Biotech and labs |
fw-19 | Attorney-client privilege and professional conduct rules | legal doctrine and professional rules | Protect confidential client communications and bind lawyers to confidentiality duties. | Legal |
fw-20 | Contractual confidentiality and NDAs | contract terms | Customer, supplier and partner agreements often limit reuse of the data you hold, whatever the privacy laws say. | Every industry |
fw-21 | Trade secret protection | law and contract | Protects information that has value because it is secret, and can be lost if it is shared carelessly. | Manufacturing, semiconductors, aerospace, professional services |
Data Compliance by Industry
A plain-language map of what to look for, and which rules to ask your counsel about, before licensing company data to an AI lab. It covers 21 industries, 21 rules and standards, and 170 industry keywords: the record types, field names and terms that signal sensitive or restricted content.
Use it to start a review, build a scanner's keyword list, label records by risk, or brief your counsel. It does not make anyone compliant, and it is not a list of what the law requires. Whether a rule applies depends on your data and where the people in it live.
Three tables
industries (21 rows): one row per industry.
| Field | Meaning |
|---|---|
industry |
Logistics and freight, Construction and trades, Professional services, Agencies, Accounting, Legal, IT services, Staffing and recruiting, Software, Manufacturing, Insurance, Healthcare administration, Finance, Semiconductors, Biotech and labs, Aerospace, Property management, Retail and e-commerce, Gaming, Air charter, Wholesale and distribution |
data_typically_held |
the records that industry usually holds |
sensitive_items_to_find |
what to search for before anything leaves |
rules_to_ask_counsel_about |
list of laws, standards and contract terms that often come up |
cleaning_focus |
where de-identification effort usually goes |
keywords |
the keyword list for that industry |
notice |
the standing caveat |
frameworks (21 rows): framework, type (law, rule, standard, contract term), about in one sentence, comes_up_in.
keywords (170 rows): industry, keyword, kind.
kind |
Meaning |
|---|---|
identifier |
names a person, account or party directly |
sensitive_field |
a field or detail that can expose a person or a business |
document_type |
a kind of record that tends to carry sensitive content |
restricted_term |
a word that signals privileged, controlled or consent-bound content |
Load it
from datasets import load_dataset
kw = load_dataset("aitrainingdatabroker/data-compliance-by-industry", "keywords", split="train")
health = [r["keyword"] for r in kw if r["industry"] == "Healthcare administration"]
How to use it well
- Find your industry and read
sensitive_items_to_findandcleaning_focus. - Run the keywords over your records to see where the sensitive content sits. A hit is a reason to look, not a verdict.
- Take
rules_to_ask_counsel_aboutto your counsel, and where a rule requires a qualified expert to sign off, use one. - Clean the data, review a sample by hand, and approve it before it leaves your control.
Limits
- One broker's summary, written in October 2026. Laws, guidance and enforcement change, and each industry has more rules than are listed here.
- Keywords are plain strings in English. They miss misspellings, abbreviations and other languages, and many hits will be harmless.
- Laws and standards are named so you know what to ask, not to say they apply to you. A standard such as PCI DSS is a contract requirement, not a law.
- Not legal advice and not a certification. Using this list does not make data compliant or de-identified.
build.py rebuilds the three files from the source lists.
License and credit
CC BY 4.0. Please credit AI Training Data Broker and link to https://aitrainingdatabroker.com/de-identification/ when you share or build on it.
About
AI Training Data Broker is a broker and technical advisor for companies that sell or license their data to AI labs. It guides the technical side, including export, format, de-identification and compliance, and publishes free checklists on removing personal information and keeping data within the privacy rules that apply.
- Downloads last month
- 33