Why This Site Now Has a Library Attached to It

A dated post is not a rulebook. Treating it as one is how operators end up quoting last year's timeline as if it were still the law.

On 28 July 2026 I published a short founding note, Opening the workshop notes, that split this site into two places on purpose. This is the follow-through. The blog stays. A library now sits beside it, and the two are not the same job.

Two clocks

The blog is news from the bench. A post has a publish date. It can be wrong later without being dishonest, because it was a judgement at the time. You read it like a magazine: dated, opinionated, finished when it ships.

The library is living reference. A page has to be correct whenever someone lands on it. That is a different contract. It is why a resource will show a last reviewed date, and why that date is meant to change. If it never moves, the page is a brochure pretending to be a handbook.

Write. Date. Revise. Those are the three jobs. Mixing them is how a useful note becomes a trap.

What “last reviewed” is for

Regulation does not wait for a content calendar. The Digital Omnibus on AI, Regulation (EU) 2026/1744, entered into force on 27 July 2026 and moved some EU AI Act deadlines while leaving others where they were. A post about that argument can sit on the blog and age. A timeline of what is live versus deferred cannot. If we publish the timeline as a blog post, six months later a reader will still find it, and the date in the byline will not tell them whether the table is still true.

So the library pages that track rules carry a review cadence. Fast-moving trackers get looked at every 30 to 90 days. Explainers tied to a statute that could still shift get 90 to 180. Practice frameworks get 180 to 365. A material change (a new commencement date, a repealed Act, a code moving from consultation to force) triggers an off-cycle rewrite. The changelog on each page is the proof.

This does not apply if you wanted a feed of opinions. Stay on the blog. The library is meant to be dry.

What the library is, and is not

It is not a second blog with nicer headings, and it is not legal advice. It is the working file I would want on the desk when someone asks “does this catch us?”: structured, sourced, and dated so you can see when it was last checked.

Pages are grouped so you can enter by the problem you actually have: the EU AI Act, EU data and privacy, UK AI and data law, US and China, the practice of governing AI systems, and the practice of deciding what to automate. One flagship page pulls the live and upcoming deadlines into a single calendar. That is the bookmark.

I will still write on the blog when there is an argument to make. The library is where the argument has to survive contact with next quarter's Official Journal.

Where to start

If you want the current position, not the commentary:

EU AI Act. Timeline and current status. What counts as high-risk under Annex III. GPAI model obligations.

EU data and privacy. Where GDPR and AI overlap. Automated decisions under Article 22. International transfers for AI and cloud tools.

UK AI and data law. Who enforces what. The Data (Use and Access) Act 2025 and the new ADM regime. ICO AI and ADM code status.

US and China. US state AI law tracker. China's generative AI rules.

Practice. Building an AI system inventory. Human oversight by design. The AI vendor risk checklist. The twelve questions. Automation versus AI.

The bookmark. The global AI and data regulation calendar.

The founding note said the noticeboard would split. This is that split, kept. Use the library when you need the current map. Use the blog when you want to know what I think about the map.

If a page in the library is wrong, tell us. That is the point of dating it.