If you believe the headlines, European AI and Europe are on the brink of an AI breakthrough. Every week, there’s another press release about a “European champion” – Mistral AI in France, Aleph Alpha in Germany, Silo AI in Finland, the latest pan-European initiative. Add in German Telekom partnering with NVIDIA, Norwegian OpenAI infrastructure build, and it all looks like we are catching up with the US giants. But the European boardroom question looms large: Are we really building independent, sovereign and SECURE European AI, or just rebranding imported technology and feeding our company secrets to intelligence services and calling it progress? Let’s cut through the marketing, the regulatory fanfare, and the unicorn hype to get to the facts that actually matter for Europe’s future. 𝔼𝕦𝕣𝕠𝕡𝕖'𝕤 "ℕ𝕒𝕥𝕚𝕠𝕟𝕒𝕝 ℂ𝕙𝕒𝕞𝕡𝕚𝕠𝕟𝕤" - 𝕆𝕣 ℕ𝕒𝕥𝕚𝕠𝕟𝕒𝕝 ℍ𝕠𝕤𝕥𝕒𝕘𝕖𝕤? Let’s start with the reality behind the leading names: - Mistral AI (France): The poster child for French AI innovation, with strong government backing and big funding rounds. - Aleph Alpha (Germany): Promoted as the German alternative to OpenAI, focused on explainable, enterprise-ready LLMs. - Silo AI (Finland): Now positioning itself as Europe’s largest private AI lab, recently going all-in with AMD for its compute stack. - German Telekom + NVIDIA: Germany’s “sovereign” AI cloud partnership, meant to offer European businesses an alternative to the US Big Tech clouds. On the surface, these all sound pretty good, and they sound like real progress. But a closer look exposes a much more complicated, and, for the cautious board member, a worrying landscape. 𝕋𝕙𝕖 𝕀𝕟𝕥𝕖𝕝𝕝𝕚𝕘𝕖𝕟𝕔𝕖 𝕃𝕒𝕨 ℙ𝕣𝕠𝕓𝕝𝕖𝕞: 𝕎𝕙𝕠 ℝ𝕖𝕒𝕝𝕝𝕪 𝕆𝕨𝕟𝕤 "𝕊𝕠𝕧𝕖𝕣𝕖𝕚𝕘𝕟𝕥𝕪"? Few outside the deep compliance and infosec world are talking openly about this: Any European AI company headquartered in France or Germany is directly subject to local intelligence and surveillance laws. - In France, the Intelligence Act (“Loi Renseignement”) gives state agencies sweeping powers to require data access from local tech companies, even in the name of “national security” (see itnews). - In Germany, the new Federal Intelligence Service Act (BND law, as amended in 2022) also mandates cooperation and data sharing with the intelligence agencies when required, allowing state trojans and telephone hacking, among other “perks”, completely without cause (see freiheitsrechte).

Hallucinations are a predictable byproduct of the way models are tested and rewarded during training. Models guess rather than admit uncertainty because testing in training rewards accuracy - not honesty. This report is a milestone because it shifts hallucinations from a mystical flaw that's something inevitable and poorly understood to a solvable engineering problem. It means LLMs can be trained to say "I don’t know", if, and only if, we reward them for it. Now that we know why, 𝘸𝘦 𝘤𝘢𝘯 𝘢𝘤𝘵𝘶𝘢𝘭𝘭𝘺 𝘥𝘰 𝘴𝘰𝘮𝘦𝘵𝘩𝘪𝘯𝘨 𝘢𝘣𝘰𝘶𝘵 𝘪𝘵. https://cdn.openai.com/pdf/d04913be-3f6f-4d2b-b283-ff432ef4aaa5/why-language-models-hallucinate.pdf

Today it’s tariffs and trade threats - tomorrow, who knows? Europe’s entire digital stack still runs on American platforms. True digital sovereignty means Europe isn’t left exposed, no matter who’s in the White House. It’s not about fear, it’s about realism. What happens if the political climate shifts and "business as usual" is off the table? Are we ready for that day? https://www.euronews.com/my-europe/2025/09/06/trump-threatens-retaliation-after-eu-hits-google-with-antitrust-fine

Here's just a few: "Curso que te ayuda a ser consciente de la importancia de un buen prompt." "This course is a great starting point for someone with zero technical or AI background" "The content is clear, concise, and well-organized, making practical application easier." "enjoyed it. learned a lot. Grateful it was free" 👇 Share with your team, great way to upskill the entire crew - for free! https://www.norai.fi/courses/prompt-engineering-mastery-from-foundations-to-future/

As a European, I can't help but be amazed by the sheer ignorance the European builders pose across the industry. Without throwing stones or pointing fingers (as it is important to support and encourage, rather than look for flaws), it seems that the average Joe is in his 20s, watched a few videos and is thinking "how hard can this be", without acknowledging the compliance wall in Europe. GDPR and the EU AI Act are not jokes or something to bypass because everyone else does so too. The average European builder doesn't seem to realise that "everyone else" are not European, and hence, are not bound by the same laws. European companies cannot risk their business by breaking the law, and most of the API spaghetti wrappers do exactly that. Then the average European builder wonders why no one buys, without realising that while there's possibly nothing wrong with the offer and build, it's the market that's wrong. European companies will not risk permanent reputation loss and possibly millions in fines just to automate document handling and save a few hours. It's simply not worth the risk. Another chapter is the South Asian builders who think they can enter Europe with their own set of rules and build in Europe what they're building in South Asia. Newsflash: Epic Fail. Not knowing the market is calling for trouble. These builders won't last a year.