AI News: The most important developments of the past week (July 20, 2026 – July 26, 2026)

Dear Community,

we receive news about new tools, security updates, and regulatory changes in the field of AI almost daily. For medium-sized businesses, it is often a challenge to keep track of it all and separate the wheat from the chaff.

This is precisely where we come in. As your bridge between science and practice, we filter the week's news and present you with the developments that are truly crucial for your competitiveness.

1. Transparency becomes mandatory: EU Commission publishes rules on AI labelling

On 20 July, the European Commission published the final guidelines on the transparency obligations under Article 50 of the AI Act. They explain how to implement a requirement that applies from 2 August 2026 – regardless of the size of a company.

At its core, this covers two things. First, people must be able to recognise when they are talking to an AI rather than an employee – for example in a chatbot on a website or with an AI-supported telephone assistant. Second, AI-generated or AI-modified content must be labelled as such, meaning artificially created images, videos or texts.

For many companies this is more relevant than it first sounds. Anyone already using a chatbot, a voice assistant or automatically generated content should now check whether the necessary notices are in place. Importantly: unlike the stricter rules for so-called high-risk systems, which were recently postponed, this date has not been moved. The transparency obligation is a separate layer and applies as planned.

There is one small relief: for the technical, machine-readable marking of AI content, systems already in use before 2 August have until 2 December 2026. The basic disclosure that customers are interacting with an AI, however, applies without this grace period.

2. Data protection without compromise: Microsoft brings Mistral to Europe

On 21 July, Microsoft and the French AI provider Mistral significantly expanded their cooperation. Mistral's models are being firmly integrated into Microsoft's enterprise platforms Azure and Copilot Studio, and Mistral is building additional computing capacity in Europe.

What is really interesting is how flexibly companies will be able to operate this AI in future. Alongside the classic cloud, there is the option of using AI in a self-controlled environment or even running it completely disconnected from the internet, intended for particularly sensitive data. The data remains in Europe.

This is precisely where the benefit for mid-sized companies lies. Data protection is the biggest hurdle for many businesses getting started with AI. With European data storage and the option of running models in-house, the question shifts: it is no longer whether AI is compatible with your own data protection requirements, but only in which operating model. Anyone already working in a Microsoft environment can integrate European models in future without leaving their familiar system.

3. Performance on demand: Anthropic releases Claude Opus 5

On 24 July, Anthropic made its new flagship model Claude Opus 5 generally available. Beyond raw performance, it brings a feature that makes a noticeable difference in everyday work: a dial for computational effort.

In concrete terms, you can set per task how much thinking effort the model should apply – low, medium or high. A simple request runs cheaply in low mode, while a complex analysis gets full power. The price stays the same as for its predecessor, and according to Anthropic, Opus 5 comes close to the considerably more expensive top model on many tasks.

For companies, this is above all a cost argument. One of the most common criticisms of AI in productive use is that costs are hard to calculate. A dial like this makes them controllable: you only pay for full computing power when a task genuinely requires it. That makes it easier to build AI into existing processes in an economically sensible way.

Our conclusion: This week was less about spectacular new capabilities than about three questions that decide the success of AI in mid-sized companies: rules, data and costs. The new transparency obligation creates clarity about what has to be labelled. The Microsoft-Mistral partnership shows that data protection and capable AI can go together. And a model like Claude Opus 5 makes costs more predictable. Taken together, AI is becoming more mature and more calculable for companies.

As always: technology is the lever, but strategy decides success.

Fiona & Maureen | Tailor-made AI Consulting