Twenty-four binding and proposed AI instruments, published as Law entities for the Add AI laws bounty. Each row below is the live entity. Every one carries a link to its official text — no news stories — plus its status, its identifier, and compliance obligations written to name who they bind, with the article number.

In effect 17 · Pending 2 · Passed One Chamber 1 · Expired 1 · Withdrawn 1 · Enacted 1 · Ordered to be reported 1

These interlock rather than stand alone. The Generative AI Measures delegate their labelling standard wholesale to the Deep Synthesis Provisions and carry no real-identity duty of their own; Order No. 21, from 2026, requires its providers to file algorithms under Order No. 9. Names are English throughout; the original gazette number stays in each instrument's identifier, since that is what a reader needs to find the source document.

The AI Act anchors these, with its risk tiers, its 10^25 FLOP systemic-risk threshold and fines reaching 7 percent of worldwide turnover. Note its Annex III high-risk obligations were pushed to 2 December 2027 by the AI Omnibus. Alongside it sits one instrument that failed — withdrawn by the Commission for "no foreseeable agreement".

Federal, state and city; enacted alongside pending. Montana's is a computational-property statute carrying an AI risk-management duty, not an AI-governance act. Colorado's entry is the 2026 replacement statute — the earlier Colorado AI Act was repealed and reenacted before its duties ever took effect, and the new obligations begin on 1 January 2027.

South Korea's framework act has been in force since January 2026. The rest answer what a list of enacted laws cannot: Brazil's bill passed one chamber, Canada's died with its parliamentary session, and the Council of Europe treaty — widely described as in force — is not, on one ratification against a threshold of five.

Every official source was opened and read, not summarised from commentary.

Six researched candidates were dropped rather than padded in. Three because artificial intelligence appears in one article of sixty-four, one of fifty-six, and none of fifty — a general law with an incidental AI clause is not an AI law. Three because their legislatures' websites refuse connections, and nothing is cited here that was not opened.

Three circulating claims were checked against primary registers and refused, including the widely repeated account that the Council of Europe convention entered into force in November 2025. The depositary's own chart records one ratification against a threshold of five, and one of the three states named as having ratified has not even signed.

The set was then reviewed adversarially, and again by four independent models, and corrected against both. Those reviews found a bill whose committee substitute had replaced the text published here, instruments tagged as enacted that were not, and regulations whose stated authority was the legislature that passed them rather than the body that administers them.

Where the reviewers were wrong, they were checked and refused. One recommended adding the Colorado AI Act; it had been repealed and reenacted, so the replacement statute appears here instead. Another asserted a treaty had entered into force on a list of ratifications that the depositary's own register contradicts.