Notes on AI that actually helps.

What we learn building AI tools for real businesses — written in plain language, with real examples. No hype, no names.

Business & compliance

21 July 2026·5 min read

AI in recruiting: what changes once CV screening becomes a regulated high-risk use

An algorithm that ranks or filters candidates is not a convenience feature under the EU AI Act — it is classified as high-risk, with obligations that follow. What that means in practice, and what the research says about the bias claims recruiting-AI vendors make.

Read →

Business & compliance

21 July 2026·5 min read

Can an AI agent actually reason like a lawyer? What a 162-task benchmark found

Regulatory watch agents are increasingly pitched as able to "read the law for you." A large benchmark built with legal professionals shows where that holds up — and where it does not.

Read →

Business & compliance

21 July 2026·5 min read

Shadow AI: what a 48,000-person global study found about employees using AI without telling anyone

Employees pasting company data into a free public chatbot, unrecorded and unapproved: shadow AI is not a fringe risk. A global study run with the University of Melbourne puts numbers on how widespread it already is.

Read →

Business & compliance

21 July 2026·5 min read

Prompt injection: the security risk that comes with giving an AI agent access to outside content

An agent that reads an email, a web page or a document to complete a task is also reading whatever instructions someone slipped into that content. Prompt injection is the security problem that comes with agent autonomy — and it has no complete fix yet.

Read →

Business & compliance

21 July 2026·5 min read

Why an agent should see less data, not more: minimization as a design rule, not a compliance afterthought

The instinct when building an agent is to give it broad access "just in case it needs it." Research on what large models retain from what they process shows why that instinct is exactly backwards.

Read →

Business & compliance

20 July 2026·5 min read

AI at work: French courts keep suspending rollouts over missing works-council consultation — what it means for your SMB

On May 21, 2026, the Paris Court of Appeal confirmed the suspension of an internal AI writing assistant and company-wide ChatGPT use for lack of prior works-council (CSE) consultation, calling it an “manifestly unlawful disturbance.” Two more 2026 rulings build the same line. Here is what actually triggers the obligation, and the checklist before you roll out an AI tool.

Read →