Briefs

Denmark calls the CPR number confidential and uses it to identify everyone. What a security professor calls the register's largest breach was caught by an invoice.

I use a language model to draft these briefs. Writing is mine when the idea, the verdict on what is true and the decision to publish were mine to get right.

A recommendation feed is not trying to entertain you. It is trying to keep you watching, and it drifts towards whatever version of you is easiest to hold on to.

On 12 June 2026 the US ordered Anthropic to cut off Fable 5 and Mythos 5 for every foreign national. The sovereignty risk we kept calling theoretical just happened.

In Denmark, AI data centres have queued for more power than the entire national grid can supply, with ten-year waits and almost no jobs. Do we actually need this?

Cloud and AI dominated every booth at V2 Security 2026. Both move trust somewhere you cannot inspect, a strange direction for the security industry to take.

We demand systems that never lose data, load instantly, and obey every law. Deletion fails not because companies are evil, but because we asked the impossible.

I am not against all surveillance; CCTV recording crimes provides accountability. What I oppose is mass tracking, profiling and systems that judge us blindly.

VPNs solve real problems, but needing them signals deeper failures. We use them to patch insecure networks, surveillance capitalism, and artificial access barriers.

Modern software runs thousands of packages from strangers. Run npm install and you run code from people you never met. Supply chain attacks exploit that trust daily.

The EU Chat Control proposal scans everyone's private messages before sending, shifting from targeted investigation to pre-emptive monitoring of all citizens.

CPU clock speeds stopped scaling in 2005 when Dennard scaling broke down. The real bottleneck is moving data, not computing it, so hardware got specialised.

Digitalisation turns public records from manual lookups into mass extraction, exposing your name, address and finances to anyone willing to query them.

LLMs generate valid code fast, but they cannot tell a secure pattern from an insecure one that looks identical, because they do not know what your code is for.

Twenty years of digital convenience built detailed profiles of our health, finances and relationships, then sold them to advertisers and governments on request.