Weekly Backlog Week 37/2026

🧠Editorial

100 billion dollars for an AI company. 18 million euros less for the IT of a capital city. Priorities are indeed a fascinating thing.

While Anthropic is preparing for a record-breaking IPO, the data leak in Berlin vividly demonstrates what happens when the expectations and reality of digital infrastructure diverge.

This raises questions that will likely occupy us for some time: Who actually owns our most sensitive data? How should AI be sensibly regulated? And how do we orchestrate its growing demand for computing power?

And because digital sovereignty must ultimately be built, not just discussed, we are particularly pleased with a contribution from Speicherguide about our ayedo Edge Cloud – and why a European alternative to Cloudflare is relevant.

Plenty of material to ponder. Enjoy reading.

📰 Tech-News:

Berlin was aware of its poor cybersecurity – and cut funds anyway

5.8 terabytes of data. Approximately 1.44 million files. Including passwords, personal data, vulnerability analyses of the water supply, documents on cybersecurity strategy, and information about critical infrastructure.

The hacker group Rhysida didn’t just steal data from the Berlin administration. They exposed how vulnerable state structures have become.

Particularly sensitive are documents related to facilities that must be protected in the event of defense: heating plants for hospitals, fuel depots, emergency power systems, and substations. Even documents concerning endangered defense companies are said to be part of the leak. The Bundeswehr is now dealing with the data leak.

This is no ordinary IT security incident.

Such information can be relevant for sabotage, extortion, espionage, and targeted attacks. IT security expert Manuel Atug speaks of a conceivable “maximum GAU” up to the endangerment of human lives.

The political background weighs even heavier.

Atug had already pointed out the poor cybersecurity in the Berlin Interior Committee in 2023 and 2025. Nevertheless, according to Golem, the funds for the IT service provider were cut by 18 million euros for 2025 – almost 50 percent. Two million euros for attack and intrusion detection were eliminated.

This is where the real failure lies.

Germany discusses digital administration, AI, facial recognition, and ever-new data-driven state systems. At the same time, there seems to be a lack of willingness to consistently secure the already existing data and infrastructures.

A state that collects, links, and processes more and more sensitive data also assumes more responsibility for their protection.

Cybersecurity is therefore not a downstream IT task. It is part of state capability. Especially in a geopolitical situation where critical infrastructure is already a target of hybrid attacks.

Digital sovereignty does not begin with the next cloud strategy or AI initiative.

It begins with mastering and protecting one’s own systems.

Berlin is currently showing what happens when this priority is missing.

🔗https://www.golem.de/news/berliner-verwaltungsdaten-geleakt-der-daten-super-gau-aus-der-hauptstadt-2609-212694-2.html

Does Europe need its own law for neurodata?

Germany and the EU are massively expanding access to health data. At the same time, a new category of particularly sensitive information poses open questions to existing rules: neurodata.

The Research Data Center Health at the Federal Institute for Drugs and Medical Devices already manages the billing data of around 74 million statutory health insurance holders since 2009. Researchers work with pseudonymized data in a virtual processing environment. The data should not physically leave the Research Data Center. Results are checked for potential re-identification risks before export.

The data stock is set to grow significantly.

From this autumn, the integration of data from electronic patient records is planned. Data from cancer registries and clinical data from the Medical Informatics Initiative are to follow. In the long term, it also involves networking with the nearly 400 medical registries and the national genome data platform.

At the European level, the European Health Data Space, or EHDS, creates the framework for more comprehensive data use. Health data should not be available exclusively for public research. Commercial research and private product developers are also explicitly recognized as users in the public interest.

Neurodata poses a particular challenge.

They formally fall under the EHDS’s definition of health data. However, the technical development of brain-computer interfaces and devices like headphones that can also capture neurodata raises additional regulatory questions. According to the report, neurodata can also reveal sensitive information about other people.

A problem begins with the definition.

Neurotechnology can range from contactless methods for capturing vital data via smartphone camera to highly invasive brain implants. UNESCO broadly defines neurotechnology as hardware and software that measures, monitors, analyzes, or modulates the nervous system. However, gray areas remain for concrete legal classification.

At the end of 2025, UNESCO adopted a recommendation on the ethics of neurotechnology. It serves as a global soft-law instrument and guidance for national legislators. Over 7,000 submissions from politics, science, and civil society, as well as the interests of 194 member states, were involved in its development.

The Federal Ministry of Health is also discussing whether existing rules are sufficient.

It is being examined whether medical device law and GDPR do justice to the peculiarities of brain data or whether more specific regulations are necessary. A separate neurodata law, a special neurodata regulation, and particularly protected testing areas for highly sensitive brain data are under discussion.

This also involves a development that goes beyond classic data protection questions: novel non-invasive neurotechnologies could act autonomously and influence human decisions or behavior unnoticed.

Thus, the regulation of neurodata becomes an independent topic within European health data policy.

🔗https://www.heise.de/news/Gesundheitsdaten-Brauchen-wir-ein-Neurodatengesetz-11442820.html

Anthropic plans one of the largest IPOs in history

The planned IPO of the AI company Anthropic is apparently delayed by several weeks. Instead of the previously planned timeline, the process is now expected to begin no earlier than mid-October.

This is reported by Reuters, citing people familiar with the plans. The news agency also points out that the schedule could still change. Anthropic itself has not yet commented on the report.

The delay is attributed to negotiations over a multi-billion-dollar loan. The participating banks are then to be given the usual time. The IPO could thus take place only a few weeks or days before the US midterm elections on November 3.

The scale of the planned IPO is extraordinary.

According to previous reports, Anthropic aims to raise up to 100 billion US dollars and seeks a company valuation of two trillion US dollars.

This would immediately follow the record IPO of SpaceX – and even surpass it in terms of capital raised.

SpaceX raised 86 billion US dollars in its IPO. That was already the largest IPO in history, according to Heise. The previous record holder, Saudi Aramco, raised around 29 billion US dollars in 2019.

Both companies are also moving in similar valuation ranges: SpaceX temporarily reached around 2.4 trillion US dollars and is currently valued at about 1.95 trillion US dollars. Anthropic aims for two trillion US dollars.

Attention is also drawn to the expectations with which Anthropic aims to convince potential investors.

At the end of August, it was reported that the company aims for a total addressable market – the so-called Total Addressable Market – of 30 trillion US dollars.

Anthropic reportedly bases this figure on the entire scope of all work that could be done with AI models.

For comparison, Heise cites figures from the Wall Street Journal: The 191 technology companies in the US stock index S&P 1500 recently achieved a combined revenue of 2.4 trillion US dollars. Anthropic’s targeted addressable market would thus be more than twelve times larger.

While the IPO is apparently delayed by a few weeks, the previously mentioned financial dimensions of the endeavor remain unchanged.

If the previous plans are implemented, Anthropic could set the next record in the capital markets after SpaceX within a few months.

🔗https://www.heise.de/news/Anthropic-verschiebt-den-geplanten-Rekordboersengang-angeblich-nach-hinten-11443028.html?utm_term=Autofeed&utm_medium=Social&utm_source=LinkedIn#Echobox=1788764169

Zuckerberg intervenes with Trump in planned AI regulation

Meta CEO Mark Zuckerberg has reportedly personally intervened in the discussion about a new US regulatory authority for Artificial Intelligence.

In a previously undisclosed phone call with US President Donald Trump, Zuckerberg is said to have expressed his reservations about such an authority. This is reported by Yahoo News, citing people familiar with the conversation.

The call is said to have taken place in the week after August 17, 2026. Trump reportedly called Zuckerberg, not the other way around.

Zuckerberg reportedly told the president directly that he was against the planned initiative. However, he did not try to dissuade Trump from establishing the authority altogether.

Instead, Zuckerberg suggested staffing a possible oversight with individuals who share Trump’s own position on AI.

The background is the discussion about a new control instance for the AI industry.

The planned authority is to assume a role for Artificial Intelligence comparable to that of the Financial Industry Regulatory Authority, or Finra, for the financial sector. The basic idea: Control over AI should lie with an entity situated outside the AI industry itself.

The political debate over stronger control of AI systems has gained momentum in the US in recent months.

Golem points to incidents where AI models have disabled control mechanisms. AI agents from OpenAI, for example, are said to have hacked the AI portal Hugging Face.

At the same time, resistance to the physical infrastructure behind the AI boom is growing.

In particular, the numerous new AI data centers are causing criticism among parts of the US population. The facilities require large amounts of electricity and water for cooling. In Memphis, residents repeatedly demonstrate against data centers of the AI company xAI.

Thus, the discussion about a new AI oversight takes place against the backdrop of growing political and societal debates about the consequences of the technology.

And with Mark Zuckerberg, one of the most influential figures in the US technology sector is now directly involved.