aiguide.substack.com — 14/01/2026

I always appreciate this author's talent for breaking down her research for laypersons like me. This article gives an overview of the current state of evaluating the cognitive capabilities and the shortcomings of these methods. Most importantly, it also provides suggestions for improving them. Definitely a recommendation for enthusiasts, sceptics and everyone in-between!

cryptography.io — 14/01/2026

Another shaky foundation of the internet: OpenSSL is an Open Source cryptographic library. It is used by most programs on the internet for cryptographic operations, including, for example, setting up a secure internet connection and encrypting emails. Like many other Open Source projects, the people behind the project are not paid, so there is a long list of bugs and inefficiencies, which is only growing over time. Alternatives exist, but they are not widely used.

The Guardian — 12/01/2026

This article highlights the vital role of governments and regulators in ensuring responsible digital innovation and protecting citizens—especially vulnerable groups—from AI-driven exploitation. It also raises the issue of digital sovereignty, showing that regulation is not just technical or legal but shaped by geopolitical pressures. For instance, action against the nudifier tool on Musk’s X platform is complicated by concerns over U.S. reactions, illustrating how global politics influence digital safety.

Rest of World — 08/01/2026

This article caught my attention for several reasons. Firstly, Russia's surprising absence from the AI race. Secondly, there is the strategic positioning of Middle Eastern players, such as the UAE, who are using their immense investment capabilities to manoeuvre between superpowers. Thirdly, this is not just about AI; it's also about broader digital sovereignty. Unlike the distinct blocs of the Cold War era, today's landscape comprises intertwined, overlapping networks in which countries pursue polycentric strategies.

Wired — 09/01/2026

OpenAI is asking third-party contractors to upload real work assignments from their jobs to create human performance baselines for evaluating its AI models. While this approach holds substantial technical promise, it also comes with major legal, ethical, and governance risks, including the risk that contractors may violate their previous employers’ nondisclosure agreements or expose trade secrets.

Wired — 19/12/2025

This article is interesting because it links digital trust to systemic AI dangers—not just small tech glitches. It predicts how in 2026 AI might spread lies, spy on people, or disrupt jobs and markets faster than today's content‑moderation and governance mechanisms can manage. This leaves us to question who should control key AI technology.