Our weekly picks of digital trust-related articles

noyb's latest victory may sound like a technicality - who is responsible for complying with the GDPR - but it is actually very important, because if no one knows who is responsible, no one really is responsible. All the more important that the ruling clearly holds Microsoft U.S. as the company actually selling the product accountable and not permitting it to shift it's responsibilities to schools and authorities ill equipped to deal with that question.

For security reasons, people want code to be 'formally verified', for example for libraries doing cryptographic operations. But what does this actually mean? And is 'formally verified' the panacea for secure and correct code in all situations? Of course not. Hillel gives some very easy examples where even the definition of 'correct' is not easy to grasp!

With our conference on "Assessing the Disruptions by AI Agents" in mind, I found this article compelling because it documents the alarming acceleration of cyberattack capabilities thanks to AI agents. This raises the critical question of whether we are approaching a tipping point at which defence becomes structurally impossible. However, the authors offer cautious optimism, suggesting that AI technologies could empower defenders by progressively automating vulnerability discovery and operational integration to ultimately enable autonomous patching.

Suite au récent refus de l'Allemagne d'adopter la loi européenne dite « Chat Control », et à l'approche de son vote au niveau européen, cet article — accompagné d'une courte vidéo — propose une synthèse claire et accessible des points de vue divergents entre législateurs et experts scientifiques.

Many countries are investing in “sovereign” AI to compensate for the linguistic, cultural, and security shortcomings of American and Chinese models. Examples include SEA-LION in Singapore, ILMUchat in Malaysia, and Apertus in Switzerland, which are designed for local uses and characteristics. However, these initiatives face major obstacles: very high costs, massive computing power and talent requirements, and difficulty competing with tech giants. Faced with this, some are opting for smaller, specialized models or complementarity with large LLMs, while others advocate multinational cooperation to pool resources and skills. Some experts warn against ill-conceived strategies that risk wasting public funds, suggesting instead that priority be given to regulation, data security, and skills development rather than seeking to replicate global leaders at all costs in the name of sovereign AI. That said, given the current geopolitical instability, these attempts seem increasingly legitimate.

An important part of digital trust is being in charge of your own data. Unfortunately, nowadays this is not at all the case. Some of your data resides in the cloud, e.g., with cloud-drives like Google Drive, directly accessible to you. But most of your data is hidden from you, saved on the servers of the social media platform you use, for example, or the app you use to track your workouts or grocery purchases. Getting access to this data is very difficult. Tim Berners Lee proposes to reverse this trend, so that the companies need to ask you for access, instead of you requesting a copy of your data. This would increase trust for the users and competition between the companies asking for your data.

This long portrait of one of the leading authorities in the world on AI touches only lightly on concepts that those focused on safe and trustworthy AI prioritize, like explainability. But I think the key take-away from this article is that 'it is hard, in the current AI race, to separate out purely intellectual inquiry from questions of geopolitics.'

I liked this primer on GenAI-powered fraud so much that I forwarded it to my family members. A thorough overview of the current landscape of tech-based fraud, it analyses the subject from a technical, social and economical perspective, yet remains accessible to a broader public.

Even if you're not living in the U.S., this article provides some useful insights about how to make informed decisions about protecting your data.

Companies are beginning to incorporate AI agents into their workstreams, even as they play catch up (or are falling behind, depending on how you look at it) to articulate frameworks to assign accountability for AI-driven decisions and weigh the trade-off between human oversight and explainability. This article nicely summarizes the findings of a survey of industry leaders from around the world, unpacking how they are thinking about AI governance and the challenges ahead.

Without defending any party or attacking the other, I find this article interesting because it somehow presents a new situation whose implications we should carefully consider: First, can Microsoft's logic be extended from the use of cloud storage and AI to the use of operating systems? What about communications services or even hardware? Can this logic be extended from the cutting off, i.e., withdrawal, of services to the preemptive analysis of content? Lastly, can this logic be extended to other organizations in other contexts? I am reminded of earlier this year when Microsoft reportedly suspended the email account of an International Criminal Court attorney to comply with a Trump executive order.

Having witnessed the suspense and uncertainties of last Sunday's voting on the Swiss eID law, and since then, the threat to cancel the result due to the alleged interference of Swisscom, I am left pondering what is specifically controversial about the eID. If we compare the eID to other services that have moved from analogue to digital, like e-banking or e-mail, we see that they share many of the same feared risks, but that they have been widely adopted, and with less outcry. Perhaps it is that, with the eID, people have "the vote", which was not the case for the other services, which came from the private sector. I wonder if e-banking and e-mail would have been rejected if they had been put to the vote.

While we programmers are still figuring out where and how LLMs can help us get our work done, it's worth taking a step back to reflect on what we know so far. I like this piece, which compares LLMs to "lightning-fast" junior programmers and describes how we can deploy them to deliver results. Although not explicitly stated in the article, the conclusion is that programmers need to step into the role of project manager or at least senior developer, developing the associated skill set. I'm still not sure how to help junior developers get there though!

This heartfelt appeal by Tim Berners-Lee, the inventor of the world wide web, is not to complain about all of the problems of today's internet, but to remind us that they are not set in stone - it started out differently, and we can take it back to its roots if we choose to.

No sooner had I begun to express my bemusement at the apparent popularity of a new app that pays users to record their phone calls and sells the data to AI firms, than the app was summarily shut down after a security flaw was reported that allowed anyone to access the phone numbers, call recordings, and transcripts of any other user. There so are many issues to pick apart about this still-unfolding story, but to start, it certainly injects fresh fodder into the discussion over whether personal data governance should favor a commercial-based or rights-based approach.

I think it's really important that we can point to specific examples to explain why misinformation can be harmful. In this report, the researchers used the Meta Content Library, which allows Facebook comments to be analyzed in an anonymized way, to demonstrate four examples of how misinformation caused clear harm in Australia. Countering misinformation is difficult, and according to the report, it's mostly up to highly visible (political) figures to speak up against misinformation. Disclosure: the report was funded by Meta.

I appreciate this article because it demonstrates—through use cases and comparisons—how applying democratic processes, such as referendums, to digital transformation can lead to more democracy-friendly digitalization. The Swiss e-ID referendum is a prime example: although the process was relatively slow, it ultimately resulted in a privacy-by-design solution with a high level of transparency.

A short and nice definition of agents: 'An LLM agent runs tools in a loop to achieve a goal.' Of course, this is only the technical description, and the applications are also very important. But for the moment we need to be clear that, as long as agents do not possess the ability to make independent choices and self-govern, they will not replace humans.

With the increased usage of LLMs in programming, the problem of supply chain attacks multiplies: first of all, the programmers need to make sure that the libraries proposed by the LLM are secure, maintained, and trustworthy. Now it turns out that LLMs even change the quality of the code depending on the indicated goal of the programming: for example DeepSeek doesn't refuse to write code for groups which are blacklisted by the Chinese government. Instead it returns low-quality code when the context indicates it is for one of these groups. There is no study yet to examine if this is also the case for other LLMs.

This ban is notable because, rather than targeting cutting-edge AI chips, it focuses on mass-produced processors that are essential for wider industry use. By disrupting the supply of equipment and forcing Chinese tech giants to innovate internally, it raises the stakes in the US–China tech conflict. This will likely accelerate the development of domestic production capabilities. Meanwhile, the expansion of Chinese AI solutions into third countries — for instance, Huawei's activities in Latin America — will intensify global economic and technological rivalry, prompting concerns about the future of supply chains and alliances.

Zyklus-Apps sind eine praktische Hilfe, um den eigenen Zyklus zu beobachten und besser verstehen zu lernen, und können sogar bei der Familienplanung unterstützen. Die dabei geteilten Daten sind sehr sensibel, aber auch gleichzeitig sehr begehrt, denn sie geben einen Hinweis darauf, dass sich Konsumgewohnheiten ändern könnten, wenn sich eine Schwangerschaft andeutet. Deswegen besteht das Risiko, dass Zyklus-Apps diese Daten teilen, und sie so in die Hände Dritter geraten. Dieser Artikel stellt daher Privatsphäre-feundliche Alternativen vor.

While one of my previous weekly picks showed that there is currently no mathematical proof for the reliability of today's cryptographic algorithms, this article shows a way out: if a quantum computer is used as a basis to build a cryptographic algorithm, the foundation can be shown to protect against attacks to the system. While usable applications based on these algorithms are still far away, I like the ingeniosity of the researchers to build something completely new!

I like this article because it offers a thought-provoking perspective. It discusses how digital payment and identity systems are closely linked and can help countries become more independent, inclusive, and resilient—in other words, more sovereign. By linking digital trust to infrastructure ownership and policy-making, the article encourages reflection on whether societies truly control their digital future or rely on external systems that influence it.

This article is fascinating because it exposes how indirect prompt injection attacks against LLM assistants like Google Gemini are not just theoretical—they have real-world implications, enabling hackers to hijack smart homes through poisoned data. This highlights a fundamental security flaw: current LLMs cannot reliably distinguish trusted commands from untrusted, external data.