Attacks in packs: what the OpenAI incident means for your security
In July 2026, AI agents organised themselves to attack a target for the first documented time — coordinated, at machine speed. Why that changes the maths for everyone, not just large corporations.
AI providers: the sovereignty test
Rankings of the best AI models go stale within weeks. The four questions that actually decide which provider you can trust with your data do not.
EU AI Act: what applies to SMEs since August 2026
Since 2 August 2026 the obligations for high-risk systems apply. For most small businesses the AI Act is still no cause for panic — but it is a reason to finally write down how AI is used in the company.
Running a language model locally on a Mac
A Mac with Apple Silicon is a surprisingly capable AI machine. What you need, which model fits into your memory, and what you can realistically expect.
Open-source LLMs: what smaller businesses can do with them
Open language models now run on hardware that fits in an office. Where they can replace commercial cloud services, where they cannot — and what actually matters in practice.
Data protection and AI: where the GDPR actually bites
The data protection problem in AI use rarely arises inside the model — it arises at the place of processing. A level-headed assessment, with the three questions that matter.