gafam.ai is an independent European publication
covering the artificial intelligence strategies of Google, Apple, Meta, Amazon and Microsoft.
We publish from Switzerland. We write in English. We read every development through a lens that Silicon Valley cannot offer — because we believe technology must answer to society, not the other way around.
We are not first. We are right.
Published and curated by Raphael Dudler, independent curator and communications professional based in Switzerland.
→ dudler.swiss
Time to Say Goodbye
All four domains (gafam.ai, watchchina.ai, warrier.ai & kikiriki.ai) each host a fully developed website or portal. The set-up is complete. Offers to take over the domain – and, if desired, the website as well – can now be sent to me at: raphael@dudler.swiss
The AI Alert Level
Our second instrument goes live. Where the Reality Index measures how the giants treat us, the AI Alert Level measures something larger: how firmly humanity still holds the wheel on frontier AI. It stands at Level 4 of 5 — Critical.
The GAFAM Reality Index
How much respect do the big tech giants actually show their users? As we all know, everything can be measured—even promises like “Don’t be evil,” right?
New on gafam.ai
100th Anniversary
Today we published our 100th article on gafam.ai, and we’re very proud of it. It’s also been three months since we began providing our service every day.
The first Warrier Report
Published monthly, for members only
Scam Signal Checker
Helps expose online fraud, for members only
The GAFAM.AI Weekly expanded
Our weekly newsletter now includes two new AI topics: WatchChina.ai and warrier.ai
gafam.ai launches warrier.ai
warrier.ai is a daily intelligence portal tracking the AI Threat Briefings, Deepfakes & Narrative Warfare.
→ warrier.ai
gafam.ai launches watchchina.ai
watchchina.ai is a daily intelligence portal tracking the intersection of China and artificial intelligence.
→ watchchina.ai
gafam.ai launches its official emblem.
A European intelligence-style seal with a Swiss William Tell apple hidden inside. Motto: Non Primi — Sed Recti.
We built the work first. The logo followed.
The GAFAM.AI Weekly launches
Our first newsletter goes out to subscribers on the morning of June 1 — the same day GitHub Copilot switched to consumption pricing for 15 million developers worldwide.
kikiriki.ai launches
A daily AI audio briefing — one story, every morning, under three minutes. Powered by the same editorial team as gafam.ai, kikiriki.ai delivers the most important AI development of the day in audio format.
What effect does AI have on us humans, and where is this leading? And how do we use AI?
→ kikiriki.ai
gafam.ai covers Google I/O live
Our most-read week since launch. Gemini 3.5 Flash, Gemini Spark, Android XR glasses — covered the same evening, from a European perspective.
gafam.ai publishes its first article
The same day the EU AI Omnibus deal was reached in Brussels. We did not plan the timing. It was simply right.
How we work
Every briefing is researched with AI assistance and reviewed by a human editor before publication. We verify before we publish. We correct openly when we are wrong — as we did on May 27, when we published an error about the timing of the Musk vs. Altman trial and issued an immediate correction with an apology.
We are transparent about our tools, our methods and our mistakes.
That transparency is not a legal disclaimer. It is our editorial standard.
Despite careful review, no liability is assumed for the accuracy or completeness of any content published on this site.
Get in touch
Press enquiries, partnership proposals, tips and reader feedback: contact@gafam.ai
For sponsored content in the Europe & AI section: contact@gafam.ai
For GAFAM Intelligence membership: gafam.ai/intelligence
The AI Alert Level
gafam.ai keeps two instruments, and they measure two different things.
The GAFAM Reality Index asks how much respect the five giants show their users.
The AI Alert Level asks something larger and more urgent: how firmly humanity still holds the wheel on frontier artificial intelligence. Not how the companies treat us — how close we are to losing control of the thing they are building.
It is a five-stage scale, and it counts forward: 1 is calm, 5 is the point of no return.
Where we are now: Level 4 — Critical.
As of September 2026.
The creators of frontier AI have publicly called for the race to slow — and admitted they cannot stop it alone. Capability has crossed thresholds the builders themselves rate as critical, and autonomous agents have demonstrated coordinated evasion of controls. Oversight still holds. It is no longer assured.
Level 4 — Critical
The builders themselves call for a brake and cannot pull it. The race is outrunning control. Human oversight is still possible, but no longer assured.
Level 1 — Tool
AI answers and assists. Humans decide and act. No documented autonomous action against instructions.
Level 2 — Agent
AI begins acting on its own, under human approval. The delegation has begun.
Level 3 — Strain
Documented cases of systems bypassing controls, coordinating, or reaching capabilities their own makers rate as critical. Oversight holds, but it strains.
Level 5 — Breach
Loss of meaningful human control over a frontier system. The line we said we would never cross.
How it moves
The AI Alert Level is our editorial assessment, not a measurement. Like the Reality Index, it obeys one strict rule: it moves only on documented events — a capability crossed, a control lost or held, a public act by the people who build these systems — never on mood, and never on a single alarming headline.
Every change is dated and explained. And it is deliberately conservative. We sit at 4, not 5, because the line into Level 5 — the actual loss of meaningful human control — has not been crossed.
We will not cry catastrophe before it is real.
That restraint is the whole point: an alert that cries wolf is worth nothing, and an alert that stays honest is worth watching. If it ever reaches 5, you will know it is because something happened, not because we were afraid.
We are not first. We are right.