Factual overview of public EU regulatory sources. Not legal advice.
EU AI Act compliance monitoring has a strange problem: most of the coverage is about a deadline, and half of what people believe about that deadline is now out of date. The Act is often discussed as if nothing is in force yet and everything lands on 2 August 2026. In reality, a large part of it has been live since 2025 — and the piece everyone was bracing for on 2 August 2026, the high-risk system obligations and the public high-risk AI database, was just deferred to 2027 by the EU’s Digital Omnibus. This post is the practical map: what you can actually track today, what changed, and how to watch the moving parts without reading 27 national websites by hand.
Two clocks, not one
It helps to stop thinking about a single deadline and start thinking about two clocks.
The clock that already struck. Several obligations are in force right now:
- Prohibited practices — banned since 2 February 2025.
- General-purpose AI (GPAI) transparency and model obligations — applicable since 2 August 2025, including the requirement for GPAI providers to publish a public summary of their training content using the AI Office template.
- Governance and enforcement structure — the European AI Office and national competent authorities have been standing up since August 2025, and the penalty provisions (fines up to €35 million or 7% of global annual turnover under Article 99) already apply to the obligations that are in force.
The clock that just got reset. The original timeline put the high-risk system obligations — conformity assessment, technical documentation, CE marking, and registration in the Article-71 EU database — on 2 August 2026. The EU’s Digital Omnibus (political agreement reached 7 May 2026, endorsed by the European Parliament on 16 June and given final Council green light on 29 June 2026) deferred those high-risk obligations for stand-alone Annex III systems to 2 December 2027, and for AI embedded in regulated products under Annex I to 2 August 2028. So the high-risk AI database does not open to registrations on 2 August 2026. That date remains a real milestone for governance and penalties, but the high-risk wave everyone circled has moved.
For a compliance team, the takeaway is simple: there is live, obligation-bearing regulatory activity to monitor today, and a smaller set of dates further out. The panic framing gets both halves wrong.
What each existing surface actually publishes
Underneath the headlines are a handful of official EU pages that publish the substance. These are the sources a monitoring workflow should watch. To make this concrete, the examples below are actual records pulled from a live run of our EU AI Act Tracker (run ID U3yNPbhBHkZhPdLDo), returned as clean JSON with fields feed, source, title, link, publishedDate, and recordType (live or roadmap-stub).
GPAI transparency (live)
This is the part that is genuinely operational. The AI Office publishes the machinery GPAI providers must use:
- “Template for general-purpose AI model providers to summarise their training content” — AI Office. The public-summary template itself, the format a provider fills in.
- “Explanatory Notice and Template for the Public Summary of Training Content for general-purpose AI models” — AI Office. The notice that explains how the template is meant to be used.
- “Drawing-up a General-Purpose AI Code of Practice” — AI Office. The Code of Practice process that many providers are aligning to.
If you provide or integrate a general-purpose model, these are the pages whose updates change what you owe.
Enforcement and governance (live)
The second cluster is the governance surface — who enforces, under what framework, on what timeline:
- “European AI Office” — the AI Office’s own page, the central body coordinating GPAI supervision.
- “AI Act” — the European Commission’s regulatory-framework page.
- “Implementation Timeline” — a maintained view of which obligations apply when (the source that reflects changes like the Digital Omnibus deferral).
High-risk registry (roadmap stub, today)
The third cluster is where the honesty matters. The high-risk obligations are not in force yet, so there is no public registration database to read. What exists today is guidance:
- “Guidelines for providers and deployers of AI high-risk systems” — AI Office.
- “Draft Commission guidelines on the classification of high-risk AI systems” — AI Office.
In the tracker these come back with recordType: roadmap-stub and a note stating plainly that the Article-71 EU central database public surface — originally slated for ~2 August 2026 — was deferred to ~2 December 2027 by the Digital Omnibus, and that the integration point activates when the central registry actually publishes. That is the correct posture: label what is guidance as guidance, and give the true date.
What changes on 2 August 2026 — concretely
Even with the high-risk deferral, 2 August 2026 is not a non-event. It is the point at which the Act’s general applicability and the national enforcement architecture are expected to be fully in place — Member States are to have designated their competent authorities and laid down penalty rules. What a compliance team should watch from that date is governance maturing: authority designations, first enforcement postures around GPAI, and any further guidance. What is not switching on that day is the high-risk registration and conformity wave — that is now a late-2027 project. Building your calendar around the deferred date, not the old one, is the single most useful correction you can make right now.
The automation angle
Here is the actual pain. This information is spread across the AI Office, the Commission’s digital-strategy library, an implementation-timeline maintained separately, high-risk guidance pages, and — as national enforcement grows — up to 27 national authorities. Checking all of that by hand, on a schedule, so you notice when a template or a timeline changes, is exactly the kind of repetitive watch that gets dropped the week you are busy.
That is what the EU AI Act Tracker is for: one run pulls the GPAI, enforcement, and high-risk feeds into a single structured dataset. Input is minimal:
{
"feeds": ["gpai", "enforcement", "high_risk"],
"maxResults": 25
}Put it on a weekly cron so a standing watch runs itself:
{
"cronExpression": "0 8 * * 1",
"runInput": { "feeds": ["gpai", "enforcement", "high_risk"], "maxResults": 25 }
}Pricing is pay-per-event: $0.01 to start a run plus $0.10 per record. A full snapshot of the current sources is eight records — about $0.81 for a complete weekly compliance snapshot, well under a dollar. The universe of records is deliberately small; each one is a source a compliance, legal, or policy team actually needs to see when it moves.
The honest summary
The EU AI Act is neither “not here yet” nor “all landing on 2 August 2026.” GPAI transparency, governance, and the penalty regime are live now and worth monitoring. The high-risk database and conformity obligations — the part most coverage still frames as imminent — moved to December 2027. Track the live surfaces, mark the corrected dates, and let a scheduled run do the watching so a template change or a timeline update does not slip past you.
This article is a factual summary of public EU regulatory sources and the current implementation timeline. It is not legal advice. For obligations specific to your organisation, consult qualified counsel.
See also: New — EU AI Act Tracker — GPAI, Enforcement & High-Risk Registry in One Feed
We cover AI agents, automation, and the tools that make them work. Our mission is to make AI accessible to everyone.