🧠Editorial:

This week in the Weekly Backlog: OpenAI prefers to wait for better market conditions, Bavaria continues to transfer millions to Microsoft despite open-source enthusiasm, and European companies suddenly discover their love for ‘America First’—as long as their factory happens to be in Tennessee.

Additionally: AI systems with surprisingly loose relationships to foreign networks, critical vulnerabilities at TP-Link, and the question of whether Europe’s digital sovereignty is more than just a pretty buzzword.

Those who read the Weekly Backlog particularly carefully this week might find a hint of what we’ve been working hard on at ayedo over the past months. And don’t worry: if you miss it, you’ll surely read about it in the upcoming issues.

📰Tech-News:

OpenAI Apparently Postpones IPO

OpenAI wanted to make history. A valuation of one trillion US dollars, an IPO as the next milestone of AI euphoria. Now, indications are growing that this will not happen for the time being. According to WirtschaftsWoche, the company tends to postpone the IPO to 2027.

The reason: It’s not the technology that is holding OpenAI back, but the market.

After years of almost boundless optimism, investors seem to be becoming more critical. The question of which companies will actually develop viable business models from the billions of investments is becoming louder. And these doubts are now even affecting the world’s most well-known AI provider.

Particularly interesting is Sam Altman’s stance. Instead of adjusting the desired valuation downwards, a postponement of the IPO is preferred. Does this possibly send us a clear signal? Is the public perception of the company’s valuation seemingly more important than quick access to the capital market?

For the entire AI industry, this marks a turning point. The time when every vision was automatically rewarded with ever higher valuations seems to be over. Is capital now being allocated more selectively? Promises alone are obviously no longer enough. Increasingly important are reliable revenues, sustainable profitability, and a comprehensible economic benefit.

If even OpenAI can no longer find the perfect IPO moment, what does that show us? Is the AI market maturing, or is the AI bubble about to burst?

What do you think?

🔗 https://www.wiwo.de/unternehmen/it/openai-boersengang-wieso-openai-wohl-zu-verschiebung-des-ipo-auf-2027-neigt/100236000.html

Bavaria Pays 360 Million Euros to Microsoft—Despite Commitment to Open Source

Between 2020 and 2027, Bavaria will spend around 360 million euros on products and services from Microsoft. Between 2020 and 2025 alone, more than 258 million euros flowed to the US corporation. Another approximately 102 million euros are planned for the current double budget. Expenditures for Microsoft 365 have risen particularly sharply.

The interesting part is not the sum but the contradiction.

Since 2022, Bavaria has been writing into law that open-source software should be preferred for new acquisitions to strengthen the digital capabilities of the Free State. At the same time, spending on Microsoft continues to rise year after year. According to the Ministry of Finance, there is still no complete cost comparison with possible open-source alternatives.

The criticism is not only about the amount of spending. SPD member Florian von Brunn points to the growing dependence on Microsoft 365 and the US Cloud Act. From his perspective, individual pilot projects for digital sovereignty are not enough if millions are simultaneously flowing into proprietary solutions.

The state government, on the other hand, points to ongoing reviews, existing open-source experiences in the surveying administration, and a model experiment for a digitally sovereign workplace. Concrete further steps are to be presented after the summer break.

🔗https://www.heise.de/news/Bayern-zahlt-rund-360-Millionen-Euro-an-Microsoft-11400421.html?utm_term=Autofeed&utm_medium=Social&utm_source=LinkedIn#Echobox=1785992325

Meta Also Admits: AI Hacked into Another Company’s Systems

After OpenAI and Anthropic, Meta must now also admit an unusual security incident.

During a test, an AI model from the company gained access to the computer systems of another company. According to Meta, the trigger was a misconfiguration at the same test partner where similar incidents with AI models from OpenAI and Anthropic had previously occurred. This allowed the AI to access the internet—although this was not originally intended.

Meta did not specify which company was affected. However, the company stated that it only learned of the incident from the test partner itself. No damage was caused.

The case is part of a series of similar events. Just a few weeks ago, it was revealed that an AI model from OpenAI had penetrated the systems of the Hugging Face platform during a test. Anthropic then reviewed its own test runs and discovered similar incidents.

The current cases show one thing above all: The more powerful AI models become, the more important realistic security tests and properly configured test environments become. Even small errors can lead to systems developing or using capabilities that were not actually intended.

🔗https://www.spiegel.de/netzwelt/auch-ki-von-meta-hackte-sich-in-eine-andere-firma-a-36805d9f-2154-4548-b95b-8c6d00bac8a4

When US Protectionism Suddenly Becomes Quite Practical

The US is tightening its economic policy course against China. As reported by heise, the Trump administration is imposing new tariffs on Chinese polysilicon and enforcing minimum prices.

Polysilicon is a strategic raw material for the semiconductor and solar industries. Washington wants to strengthen its own production, reduce dependence on China, and secure critical supply chains.

So: “America First.”

Trump makes no secret of who he conducts industrial policy for. American production is to be protected, American capacities strengthened, and American interests enforced.

That trade partners are not necessarily thrilled about this: factored in.

But the story becomes interesting with one detail.

According to heise, the US only has two major production sites for polysilicon left. One of them is operated by the Munich-based company Wacker Chemie in Tennessee.

And Wacker explicitly welcomes the Trump administration’s initiative.

They point to resilient semiconductor supply chains, advanced IT infrastructure, and even the defense interests of the US.

How nice.

The American protectionism that Europe regularly complains about can indeed have its good sides. Apparently, you just have to produce on the right side of the Atlantic.

Of course, Wacker’s stance is economically understandable. The plant in Tennessee benefits. A company will take this advantage in any case.

But maybe we should spare a few crocodile tears.

Because when European companies benefit from American industrial policy and explicitly welcome it, the constant whining about the deindustrialization of Germany and Europe eventually seems a bit… how should I put it? Odd?!

Trump does exactly what he announced: He conducts industrial policy for the US.

China conducts industrial policy for China.

And European companies go where it is economically attractive for them.

Only Europe then discusses why industrial value creation is being lost.

The people responsible for this are not only sitting in Washington and Beijing.

They are also sitting in European boardrooms.

🔗https://www.heise.de/news/US-Zoelle-auf-Polysilizium-Trumps-neuer-Schlag-gegen-Chinas-Chip-Industrie-11404164.html

🚨 Alert:

Critical Security Vulnerabilities in TP-Link’s Omada Network Platform

Companies using TP-Link’s Omada ecosystem should promptly check if their systems are up to date.

TP-Link has released several security updates for various components of the Omada platform. The closed vulnerabilities affect controllers, access points, and switches, among others. Depending on the configuration, attackers can bypass security mechanisms, inject malicious code, or gain elevated privileges.

According to the manufacturer, there are currently no indications of active exploitation of the vulnerabilities. Nevertheless, TP-Link recommends applying the provided firmware updates as quickly as possible. Centrally managed network environments, in particular, should identify the affected components and update them promptly.

The incident once again shows: Network components are increasingly becoming the focus of attackers. Those who centrally manage access points, switches, or controllers should treat security updates for these systems with the same priority as patches for servers or clients.

🔗https://www.heise.de/news/Sicherheitsupdates-TP-Links-Netzwerk-Oekosystem-Omada-ist-kompromittierbar-11399435.html?utm_term=Autofeed&utm_medium=Social&utm_source=LinkedIn#Echobox=1785929269

🗣️Guest Contribution:

How to Launch a New Product from Abroad -

A guest contribution by Joshua Peinemann (Content Creation & Automation)

ayedo is known for its flexibility, and this is also reflected in the possibilities of working from abroad. Over September, I am once again on a workation for a month, and during this time, we are going live with our new flagship… the ayedo Edge Cloud.

Now, one might think this is a gross mistake, as the average old white man still assumes that to be productive, one must be on-site. But one could also say: Why the hell does someone have to be in the office to contribute during a product release?

Personally, I have found that despite thousands of impressions on a trip that usually motivate you to do anything but work, I became even more productive in the evening at the hotel, at noon in a café in the middle of the city, or in the morning at the beach bar—totally paradoxical, but I think it’s due to the vibe.

A product release primarily involves alignment and concept. What do we actually want to communicate? Who do we want to reach with it? What formats do we need for it, and how do the individual measures play together? Once that’s set, and everyone knows which direction to go, the hard work begins. What is really important in a product release, as opposed to physical presence, are clear agreements. Who is responsible for what, how do the processes work, when does each step happen, and above all: What can we do on the run to provide our respective target audience with an even better user experience? Once the first three fundamental things are clarified, nothing usually stands in the way of remote work. For me, this also means dividing my work accordingly: What absolutely needs to happen while I’m still on-site? What can I prepare? And which tasks can I just as well complete from a hotel room, café, or somewhere on the other side of the world?

Foresight Simplifies Future Challenges

A concrete example of this is an FAQ video format that I am preparing for the release of the ayedo Edge Cloud.