The EU AI Practitioner's Handbook

A working reference for anyone building or buying AI in Europe: the news sources worth your time, the analysis that actually holds up, an EU regulatory calendar with verified dates only, and the toolbox we use to measure and build AI visibility — the discipline behind our AI Visibility service. Curated by CogniLedger — the same list we use ourselves.

Last verified: 2 July 2026

News & Signal

Ten sources, three layers: primary vendor announcements, industry coverage, and filters that save you the rest.

Primary sources
Anthropic — News

Model releases, safety research, MCP ecosystem announcements — first-hand.

OpenAI — News

Product and model announcements from the largest consumer AI vendor.

Google DeepMind — Blog

Gemini releases and research; the second model family we deploy in production.

Mistral AI — News

The EU-native frontier lab — essential reading for European vendor-selection decisions.

Industry coverage
TechCrunch — AI

Fast, broad coverage of funding, launches, and policy moves.

The Verge — AI

Consumer-angle AI coverage; good early signal on product shifts.

Filters — newsletters & blogs
Ben's Bites

The fastest way to skim what happened in AI yesterday.

Import AI (Jack Clark)

Research and policy digest by an Anthropic co-founder; unusually good on governance.

Simon Willison's Weblog

The most reliable hands-on evaluations of new models and tools, including MCP.

Latent Space

The AI engineering angle — how practitioners actually ship with these models.

Analysis & Research

Reports and data sources that survive contact with reality — the ones we cite in client work.

Stanford HAI — AI Index

The most-cited annual snapshot of AI capability, cost, and adoption data.

State of AI Report

Benaich's independent industry review — dense, opinionated, well-sourced.

Epoch AI

Rigorous data on compute trends, training costs, and model capabilities.

Interconnects (Nathan Lambert)

Deep technical analysis of open vs closed models — beyond the hype cycle.

McKinsey — State of AI survey

Enterprise adoption numbers your board has probably already read — know them too.

SemiAnalysis

The economics of AI infrastructure — why models cost what they cost.

EU Regulation Desk

Where we track the EU AI Act, GDPR intersections, and the standards that will define conformity. This is the desk we run for our own products.

EU AI Act Explorer (FLI)

The best free full-text explorer of the Act, with an implementation timeline.

AI Act Service Desk (EC)

The Commission's own implementation-support hub — timeline, FAQs, guidance.

EC — AI policy hub

Primary source for guidelines, codes of practice, and Omnibus developments.

EU AI Act Newsletter (Risto Uuk)

The single best digest of AI Act implementation news.

EDPB

Where AI meets GDPR — opinions and guidance on AI and personal data.

IAPP — AI Governance

Privacy-profession coverage of AI governance; home of the AIGP certification.

Harmonised Standards Map

Maps CEN-CENELEC draft standards to specific AI Act articles — invaluable.

CEN-CENELEC JTC 21

The committee writing the harmonised standards that will define AI Act conformity.

Regulatory Calendar

Confirmed dates only, each with a source. No rumors, no "expected Q4". Verified against official EU sources on the date shown above.

Already in force
2 Feb 2025
AI Act: prohibitions and AI-literacy obligations apply.
2 Aug 2025
AI Act: GPAI-model obligations and EU governance structures apply.
10 Jun 2026
Commission publishes Code of Practice on marking and labelling AI-generated content.
29 Jun 2026
Council gives final approval to the Digital Omnibus on AI (EP endorsed 16 Jun); publication in the Official Journal expected July 2026.
Upcoming
Jul 2026 (expected)
Digital Omnibus on AI published in the Official Journal; enters into force on the third day after publication.Source: Council of the EU
2 Aug 2026
The bulk of the AI Act becomes applicable and enforcement begins: Article 50 transparency rules (AI-interaction disclosure, deepfake labelling), innovation-support measures, and the requirement for at least one regulatory sandbox per Member State.Source: EC AI Act Service Desk
Q4 2026 (target)
First wave of CEN-CENELEC harmonised standards for high-risk AI expected under the accelerated JTC 21 procedure (incl. QMS prEN 18286, risk management prEN 18228, data governance prEN 18284).Source: CEN-CENELEC BT decision, Oct 2025
2 Dec 2026
End of the grace period for machine-readable watermarking of AI-generated content (Art. 50(2)) for systems placed on the market before 2 Aug 2026.Source: Digital Omnibus agreement
2 Dec 2027
High-risk obligations apply for stand-alone Annex III systems (recruitment, credit scoring, education, border control, etc.) — deferred from 2 Aug 2026 by the Digital Omnibus.Source: Digital Omnibus agreement
2 Aug 2028
High-risk obligations apply for AI embedded in regulated products (Annex I: medical devices, machinery, toys, vehicles).Source: Digital Omnibus agreement

Dates reflect the Digital Omnibus on AI as politically agreed (May 2026) and approved by Parliament and Council (June 2026); final wording subject to publication in the Official Journal.

Release Watch

We don't maintain a rumor list — it would be stale in a week. Instead, the live trackers we check.

Artificial Analysis

Independent benchmarks, pricing, and speed comparisons across every major model — our first stop for vendor selection.

LMArena

Crowd-sourced head-to-head model rankings; the fastest signal on whether a new release actually moved the needle.

Hugging Face — Trending

Where open-weight releases land first.

AI Visibility Desk

AI Visibility is what we practice and sell: making a company discoverable, accurate, and authoritative when AI assistants answer questions about it. This desk collects the tools we use — free checkers to see where you stand today, monitoring platforms for the long game, and the open standards the whole discipline is built on. Start with the free checks.

Check your visibility — free, in minutes
HubSpot AEO Grader

Free one-time snapshot of how ChatGPT, Perplexity, and Gemini describe your brand — sentiment, share of voice, and gaps. The fastest honest answer to "are we visible?"

Bing Webmaster Tools — AI Performance

Free, platform-native data on how your pages are cited in Microsoft's AI experiences — insider numbers, not third-party estimates.

LLM Pulse — Discoverability Checker

Free technical audit: can AI crawlers actually reach, render, and trust your site? Checks crawl access, structure, and authority signals.

Continuous monitoring — the tracker market
Otterly.AI

EU-built AI search monitoring across ChatGPT, Claude, Gemini, Perplexity, and AI Overviews; accessible entry pricing for SMBs.

Peec AI

Berlin-based tracker with strong multilingual coverage — relevant if your customers ask AI assistants in more than one language.

Profound

The enterprise/agency end of the market: deep competitive benchmarking and white-label reporting.

Semrush — AI Toolkit

AI visibility tracking bolted onto the SEO suite many teams already run — the lowest-friction option if you're a Semrush shop.

Foundations & standards — what visibility is built on
llms.txt specification

The open standard for describing your site to AI models — robots.txt's younger sibling. We ship one for every AI Visibility client (and run one ourselves).

Schema.org / JSON-LD

Structured data remains the backbone of machine-readable trust — for AI answer engines as much as for Google.

GEO — the original paper

The 2024 research paper that defined Generative Engine Optimization and measured which content changes actually move AI citations.

Awesome GEO (GitHub)

A living, curated map of the entire GEO/AEO ecosystem — tools, research, and case studies.

AI crawler docs (GPTBot)

Know the bots reading your site: user agents, IP ranges, and how to allow or restrict them deliberately.

The MCP layer — beyond being described, being connectable
Official MCP Registry

The canonical index of public MCP servers — including ours. Registration here is how assistants find you.

MCP Specification

The protocol spec and changelog; the layer where a company stops being a web page and becomes a source AI assistants can query directly.

This is the stack behind our AI Visibility service — we ran every step on our own products first. See the service →