European Tech Sovereignty, Cyber Threats & AI Governance Roundup

CYBERSECURITY

European Publishers Push Back Against Big Tech’s Use of Their Content for AI

Publishers in Europe are taking a stand against the way large tech platforms use their content to power AI services. In the space of a few days, French publishers have turned to Brussels, the UK competition authority has raised concerns about Meta, and a Norwegian court has stepped in.

French press files complaint against Google

On Thursday, 29 October, five French organisations representing the press, online publishers and digital marketing filed a complaint with the European Commission. It concerns Google’s rollout of AI Overviews and AI Mode in France. The groups are:

  • the Collective for Digital Marketing Actors (CPA)
  • the National Federation of Specialised Information Press (FNPS)
  • the Group of Online Content and Service Publishers (Geste)
  • the Union of Magazine Press Publishers (SEPM)
  • the Union of Independent Online News Press (Spiil)

Their complaint feeds into a formal investigation the Commission opened in December 2025. Brussels is examining whether Google broke EU competition rules by using publishers’ content and YouTube videos to power its own AI services. The French groups say they want to “contribute to the investigation process by providing factual information collected from French publishers and an analysis of practices observed since the deployment of AI Overviews and AI Mode”.

UK competition authority turns to Meta

The UK’s Competition and Markets Authority (CMA) has raised concerns about Meta’s new terms of use. According to the CMA, they could allow Meta “to collect content from websites of businesses and publishers that use its services, and then use it to develop and provide its own products and services, including its AI features, without those companies having a say in whether their content is used or how it is used.” The concerns were raised in particular by the News Media Association, which represents British press publishers, and by the Movement for an Open Web.

Norwegian court halts Meta’s new terms

In Norway, a court has temporarily banned Meta from applying the same new terms, which the country’s major media outlets are challenging. They say the terms would let Meta collect their articles, photos and videos, including content reserved for paying subscribers.

Why it matters

These cases raise a key question for Europe’s digital economy: who controls the value created by content once AI systems reuse it? The outcome could shape how AI services are built and how publishers are paid in the years ahead.

1)DELCAMPE A PLATFORM TO SELL COLLECTIBLE ITEMS HIT

Among stole data are the names of the customers, their personal address, and information on their transactions. Such elements could be exploited to drive a robbery campaign.

2) MORE ON THE BINANCE HACK

This Newsletter has reported on the huge cyber hi jacking at BINANCE. We can now provide precisions on the way the North-Korean criminals managed to steal almost 6 billions €. A Binance engineer was contacted on a job-search platform by a company representative – who turns out to be fake. He then went through job interviews via video call.  During one of these virtual meetings, he was asked to download something to take a test and demonstrate his skills. Among these items was a sophisticated and stealthy piece of malware. From there, the hackers used an internal AI tool to navigate the Binance IT system

3) MORE ON THE CYBER ATTACK AGAINST FRENCH AND EUROPEAN REFRIGERATORS

Eyes turn to Iran, which is suspected to be behind this huge attack. “A judicial investigation is opened by France against the head of the Iranian group Shahid Kaveh, an offensive cyber unit affiliated with the Electronic and Cyber Center of the Revolutionary Guards (CGRI-CEC), against his two deputies, and against anyone else who might be identified,” the prosecutor detailed in a statement. “The investigation has uncovered, with the support of the State’s specialized services, an attack method characteristic of this unit, the use of Farsi, and IP addresses pointing to terminals located in Iran,” the prosecutor continued.

4) NIS2 FINALLY SOON TRANSLATED INTO FRENCH LAW ?

The French National Assembly is now working on the draft law. While there seems to be a consensus no most points, observators expect the debates will once again focus on Article 16 bis, related to messaging encryption, which the government wants to get rid of. This article bans forcing encryption or messaging service providers to install ‘backdoors,’ so the state or intelligence agencies can’t have an easy regular access to suspected IT systems. While this provision is highly unpopular with some governmental services and some parliamentary groups, who cite national security, it remains supported by most reps of the Internet community and defenders of freedom. At the moment we write, it seems that the Government will finally propose an engagement to avoid trying to force service providers to install backdoors, and simultaneously will recommend the suppression of the controversial article 16bis, which could lead to the impossibility to scrutinize communication even one by one.

5) SALT TYPHOON A PERMANENT THREAT LINKED TO CHINESE SERVICES

The Chinese state-sponsored hacking group tracked as Volt Typhoon has built a highly structured network of compromised devices and rented infrastructure to conceal access to critical infrastructure and other strategic target. Not only this attack scheme is built for long, but it has replicates: Black Lotus Labs, an US cyber consultancy, three years ago discovered a new “covert network,” codenamed Silent Train, being used by the same state-sponsored hacking group, which poses a potential military disruptive threat to the United States and its allies.

This covert network is “designed to gain and maintain access to the technologies businesses rely on to manage devices, networks and critical operations,” using a carefully crafted operational framework, says a new report issued by the consultants.

The framework typically relies on attack chains comprising four to six hops across virtual private servers, SOHO routers, IP cameras and sometimes even enterprise servers, with Volt Typhoon separating devices by function, frequently discarding infrastructure after one use and also wielding apparent zero-day vulnerabilities.

The issue with such hacking systems is that they are not designed to support one raid, but to try permanently infiltrate infrastructures and build long term spying.

6) Assess our resilience in case of giant black-out

ALLIANCY reports on a discussion hold on November 10 between IN CYBER FORUM and HEXATRUST, on our resilience in case of front attack disrupting key infrastructures. Some hints:

Simulating a France without electricity to find out, before an attacker does, where our defenses fail is the idea defended by Jean-Marc Autret, EDF group’s OT CISO. The day before, France had attributed to a unit of the Iranian Revolutionary Guards the attack that disrupted the fridges of 700 supermarkets in October.

“We need to run a large-scale exercise on what a France plunged into darkness would look like,” argued the EDF official. According to him, “if we don’t test the entire chain end-to-end, we’ll never know how resilient we really are.”

At EDF, which manages a thousand industrial sites, each system is classified according to its impact, based on a logic inherited from nuclear safety, over assets spanning from 2000 to 2025. In case of an attack, the group assesses its defense lines and then makes a prognosis that helps decide whether to isolate equipment, stop a flow, or shut down a machine. “I wouldn’t hesitate to stop a system if it’s highly exposed to identified threats,” said the OT CISO.

ECA COMMENT: in case of major electricity shut down, all our IT systems and devices will be down. As in the case of a major successful cyber attack… What about also having resilience exercises to prepare that sort of situation ?

INNOVATION AND REGULATION

THE VERONA STATEMENT: COMPUTER SCIENTISTS PLEDGE FOR A STRONG GOVERNANCE OF AI USE IN COMPUTING

We here publish the introduction to the in progress statement:

AI can change how we do science.
Not what makes it trustworthy.

The Verona Declaration on the Use of Generative AI in Computing Research is taking shape. Researchers across the computing community are developing shared principles for using generative AI while preserving scientific integrity, independent judgment, and accountability.

Researchers may delegate tasks to AI systems.
They cannot delegate scientific responsibility.

GERMAN SCHOOLS WILL HAVE A SOVEREIGN COLLABORATIVE SOLUTION

Six “Länder”: Berlin, Brême, Hambourg, Hesse, Rheinland-NordWestphalen and Rheinland-Palatinate, have opted for the NEXTCLOUD platform which provides functions such as file sharing, communication services, collaborative work on documents. The platform, named KoDaS, is hosted on the German cloud IONOS, and relies on servers located in Germany.

While the system could be extended to more German schools, France could follow suite, as its pen Source based platform apps.education.fr notably gives access to Nuage, a service based on Nextcloud, which allows staff to store, share, and edit documents together.

ADVANCED TECH

BULL AND SCHWARZ GROUP TEAM UP TO PROVIDE ENHANCED SOVEREIGN CLOUD SERVICES

The French group is teaming up with the IT and digital subsidiary of the Schwarz group to build an independent and secure European AI and cloud ecosystem, covering everything from servers to software platforms and application environments.

The partners will develop a high-performance sovereign AI architecture, combining their respective expertise to build secure large-scale AI infrastructures in Europe. Schwarz Digits will provide its secure cloud infrastructure through STACKIT, a sovereign European digital ecosystem, while Bull will leverage its AI expertise and high-performance computing experience gained from large-scale projects. The servers would rely on Bull’s European manufacturing capabilities, with its Angers factory producing rack-scale AI.

Share:

More Posts