The sixteen themes, filled from binding law

About this documentUpdated 2026-09-18ShowHide

Sean McDermott, Co-Founder and CEO, UnGovr

Written by Sean McDermott (with AI assistance) using the LexLint law library, which supplied every legal instrument, status and date on these pages.

Every law named here links to its summary page on lexlint.org, translated to English (if needed) and restructured to a standard format for human and code use. Every case links to the court's or the regulator's own record where one could be reached.

© 2026 UnGovr, publishing as LexLint. The text, the figures and the theme-register file are licensed under Creative Commons Attribution-NonCommercial 4.0: share and adapt them for noncommercial purposes with credit to LexLint (UnGovr). Please contact LexLint at hello@ungovr.org to discuss commercial use. Logos and wordmarks belong to their owners.

Corpus figures as of 2026-09-18.

The Agentic AI Foundation's governance working group is extracting requirements from frameworks under the sixteen themes below. This register runs the same extraction against statutes by jurisdiction instead: what the law already says under each theme, whom it binds, and where it reaches from, with the full register as a file.

1What this is

The Agentic AI Foundation (AAIF), a Linux Foundation foundation, runs a Governance, Risk and Regulatory Alignment working group whose charter sets three deliverables: a landscape of agentic-AI policy frameworks, a gap analysis, and distilled recommendations. Its monthly reports record the method it chose: about a dozen common themes, and extraction prompts to map frameworks, standards and legislation onto them consistently. In August 2026 the group circulated a workbook to its members with sixteen such themes, each with an extraction prompt to be run against one framework document at a time, and a reference guide of thirty frameworks to run them against. The workbook is not public; the themes are summarised here in the group's own words.

The prompts share one contract: classify every provision by type, decompose compound statements into atomic requirements, keep the exact modal language, and record for each requirement an identifier, the text, the type, the obligation strength, the applicable entity, the source clause and a confidence. Two of the prompts' rules matter most: do not infer applicability beyond what the source states, and say "legal interpretation required" where legal analysis is needed.

This register runs the same contract against a different source. Instead of thirty frameworks, the input is binding law by jurisdiction: the requirement statements the LexLint software-law corpus carries for every AI, privacy, scraping and cybersecurity instrument it holds, each one already atomic, already tied to a citation, and already dated. The output keeps the group's columns and adds the two a statute needs that a framework does not: the party the duty is addressed to, and the territorial hook it reaches through. Where a requirement has been classified for what a runtime control can do about it, that tier is shown too.

The point of doing it is the one the group's own June 2026 report made: a gap analysis needs a baseline. A framework register says what good looks like. This register says what is already required, of whom, and where, so that the gap between the two can be measured rather than described.

2Method

Source. 4,756 requirement lines drawn from the LexLint software-law corpus on 2026-09-18, across the four topics that carry them: AI (1,006), privacy (1,791), scraping (1,139), cybersecurity (820). Of those, 3,979 belong to an instrument in force on that date; the rest are enacted but not yet commenced, and are kept in the register with their lifecycle band so a reader can see what is coming.

Theme membership. A line joins a theme when its instrument's obligation class is one the theme owns, or when the line's own wording matches the theme's rule. A line can sit in more than one theme, as the group's own prompts allow. The rules are deliberately generous, so a theme's count is an upper bound on what a careful reader would keep, and the thin themes are thin because the law says little there, not because the rule missed it.

Type and obligation strength. The group's six types are derived from the line's wording and, where one exists, from an independent hand classification of the same line (obligation, prohibition, scope condition, remedy note, permission). Two values the group's list lacks are added because the law needs them: a consequence (a penalty or remedy stated as a requirement line) and a permission or exemption. The modal language preserves the verbs the line uses, as the prompts' rule 3 asks.

Applicable entity. Read off the line's own wording where it names a role (provider, deployer, controller, processor, platform, employer, manufacturer). Where it names none, the entity is the OPERATOR, the party running the application, which is whom the LexLint software-law corpus binds by default.

Territorial hook. Derived from the instrument's topic and law family under a stated rule: a privacy duty reaches through the data subject's residence and the OPERATOR's establishment, a transfer duty through the destination, a scraping duty through the COUNTERPARTY's location, an AI transparency duty through where the output is used and the market it was placed on, a security duty through the OPERATOR's establishment or the placing on the market. The column is a rule applied to every line, and it is the first thing a reviewer should correct.

Party whose position triggers scope, and runtime tier. Both come from a separate hand classification of 2,589 of the lines, made for a study of what an AGENT gateway can enforce. The party is the one of six (OPERATOR, principal, provider, COUNTERPARTY, affected person, data subject) whose location or status the duty turns on; the tier says whether a control on the request path can prevent the conduct, detect it, evidence it, or never sees it. Lines the classification has not reached say so.

Line identifier. Every line carries an identifier of the form topic:instrument-code:position, which is the topic it was researched under, the instrument's code on lexlint.org, and the line's position in that instrument's requirement list. It is stable for as long as the wording is unchanged: a reworded line gets a new identity rather than a silent change under the old one. It appears at the end of each entry below and in the file, so that a row in the group's own register can cite one.

3The register in one table

One row per theme. "Lines" is every requirement line the theme's rule admits, in force or not; "in force" is the subset whose instrument binds today; "instruments" and "jurisdictions" count the distinct sources those lines come from. The last column says which of the four topics the lines were researched under, largest first, which is a fair proxy for which body of law a theme really lives in.

4What the distribution says

Four themes carry most of the binding law: data protection and privacy (1,611 lines in force), security and access controls (1,696), transparency and USER disclosure (1,447) and recordkeeping and auditability (1,290). These are the themes where a statute, rather than a framework, is the source of the requirement, and they are where a gap analysis against binding law will find the most to measure.

Three themes are thin, and the thinness is the finding. Architecture has 20 lines, testing and red teaming 9, monitoring and logging 79. Binding law rarely tells an OPERATOR how to structure a system, how to test it, or what to log in what format; it tells the OPERATOR what outcome to secure and what record to be able to produce. Those three themes are framework territory, and the frameworks in the group's reference guide are the right source for them. What a statute adds is the outcome the framework control has to serve, which is why the recordkeeping theme is large while the logging theme is small.

Use-case classification and risk tiering (521 lines) is dominated by prohibitions: the law's way of tiering is to name what may not be done at all. Human oversight (82) and action management (97) are small but spread across many jurisdictions (67 and 77 respectively), which is the pattern of a duty that arrives through privacy law's automated-decision rules rather than through an AI statute. Incident response (887) is almost entirely breach notification, the oldest and most uniform family in the LexLint software-law corpus.

5The sixteen themes

Each theme below quotes the group's own key concepts, gives the counts, and holds a sample of up to fourteen lines in force, drawn to span jurisdictions rather than to rank them, folded away until you open it. The requirement is stated first, as the LexLint software-law corpus records it; under it, the facts the group's columns ask for, then the jurisdiction, the instrument and its date, and the line identifier. The full register, every line in every theme, is the file in section 6.

6The full register as a file

theme-register.csv holds every line that falls under at least one theme, one row per line, with the theme identifiers it belongs to, the columns above, and the topic, instrument, citation, summary page, lifecycle band and as-of date. It is generated from the LexLint software-law corpus on the date in this page's byline and it will be regenerated when the corpus moves; a copy taken today is a snapshot with that date on it. The theme columns are the register's, not the corpus's: the corpus records obligation classes and wording, and the mapping from those to the group's sixteen themes is the rule stated in section 2, which the group is welcome to replace with its own. The file, like the documents, is licensed under Creative Commons Attribution-NonCommercial 4.0: use it for noncommercial purposes with credit to LexLint (UnGovr), and contact LexLint at hello@ungovr.org to discuss commercial use.

7What this register does not claim

It does not say that any line binds any particular organisation. That is the applicability question, and it is answered per deployment from the roles a system holds and where its parties are, which is what the first document in this section is about. The theme assignment, the addressee and the hook are rules applied to text, reviewed in samples, not findings made line by line. Every entry links to its source so that a reader can check the rule against the law, and the working group is the best reviewer the register could have.