← The State of Tech RSS Feed

17 juni 2026 · 13:12

G7 AI Access Framework: What It Means for Europe

G7 leaders are weighing a trusted-partner AI access framework that would give allied nations privileged entry to America's most powerful frontier models, in exchange for aligning on chip export controls and security oversight. For Europe, the proposal creates a direct tension between staying competitive on AI and defending the strategic autonomy the EU has championed for years. This episode unpacks that dilemma alongside a major European cloud outage, a six-billion-euro space investment, and a coordinated deepfake fraud warning hitting bank customers across Germany, France, and the Netherlands.

AI G7 Cloud Outage European Space Agency Ariane 6 Deepfake Fraud

Beluister deze aflevering:

Transcript

Samantha: Welcome to The State of Tech, The European Edition, Wednesday June seventeen, 2026. I'm Samantha Lawrence.

Bob: And I'm Bob Russell. Today: G7 leaders weigh giving trusted partners access to America's most advanced AI models, a major European cloud outage exposes how concentrated the market really is, Europe's space ambitions get a multibillion-euro boost, a fresh warning about deepfake fraud hitting bank customers, and to close out, two things you can actually use today: the focus app Opal and the open-source photo editor Darktable. Let's start with the G7.

G7 leaders weigh giving trusted partners access to America's most advanced AI models.

Samantha: At the G7 summit, leaders are discussing a framework that would let so-called trusted partner countries gain privileged access to America's most advanced AI models. The idea is essentially a club: allies who agree to shared security standards, export controls, and oversight rules get to use the most powerful systems from US labs, while others stay behind a wall. Washington is driving the proposal, with the UK, Japan, and Canada reportedly receptive, and the European Union sitting somewhere in the middle.

Bob: The mechanics matter here. We're talking about access to frontier models for sensitive sectors like defence, biotech, and critical infrastructure, the kind of compute and capability that today only a handful of US firms can offer at scale. In exchange, partner countries would commit to chip export controls aligned with US policy, security vetting of who uses these models, and joint monitoring for misuse. That's a significant ask, especially for governments that have spent the past two years talking about sovereign AI and strategic autonomy.

Samantha: And for Brussels, this creates a genuine dilemma. On one hand, European companies want access to the strongest tools to stay competitive with American and Chinese rivals. On the other hand, signing onto a US-led framework cuts against the EU's own push for sovereign infrastructure and the AI Act's risk-based approach. Smaller member states worry they'd be locked out entirely if the access is gated through bilateral deals with Washington rather than the bloc as a whole.

Bob: There's also the China question lurking underneath all of this. The framework, if it goes through, is explicitly designed to keep advanced AI capabilities away from Beijing, which means European firms doing business in China could face new compliance headaches. Industry groups in Germany and France are already pushing back, arguing this turns AI into a geopolitical weapon rather than a shared technology. No formal agreement has been signed yet, and officials stress this is still a discussion document, not a treaty. But the direction of travel is clear: the AI supply chain is being redrawn along political lines, and Europe has to decide which side of the wall it wants to be on.

A major European cloud outage exposes how concentrated the continent's digital backbone really is.

Samantha: A multi-hour outage at one of Europe's largest cloud regions hit thousands of businesses yesterday, knocking out everything from online banking portals to logistics platforms and government services. The root cause, according to early statements, was a botched configuration update that cascaded across data centres in Ireland and the Netherlands. Services were largely restored within four hours, but the disruption rippled through dependent systems for much longer.

Bob: What makes this story bigger than a typical outage is the concentration risk it exposes. A handful of hyperscalers run the vast majority of European cloud workloads, and when one region goes down, the effects spread across borders almost instantly. Hospitals reported delays in patient record systems, several regional airports had check-in disruptions, and at least two large retailers lost their payment processing during the morning rush. The European Commission has already signalled it will look at whether this falls under the new digital resilience rules that came into force this year.

Samantha: Those rules, known as DORA for financial services and the broader NIS2 directive, require critical operators to have tested fallback plans and to report major incidents within tight windows. Yesterday's outage will be a real-world test of whether companies actually have working backups or just paperwork. Early signs suggest a lot of organisations discovered the hard way that their disaster recovery plans assumed the cloud itself would never fail.

Bob: And this feeds directly into the sovereign cloud debate. European providers like OVH and Deutsche Telekom have spent years arguing that overreliance on American hyperscalers is a strategic vulnerability, and yesterday gave them a fresh talking point. Whether that translates into actual contract shifts is another question, because the price and feature gap remains wide. The bottom line for users: if your bank app or your delivery tracker froze yesterday morning, this is why.

Samantha: Quick interruption. If you listen to The State of Tech regularly, hit that like button and subscribe, that way you'll never miss an episode. Okay, moving on.

Europe pours fresh billions into its own space launch capacity to close the gap with SpaceX.

Bob: The European Space Agency and the European Commission announced a joint funding package worth around six billion euros to accelerate Europe's independent launch capabilities over the next five years. The money will be split between scaling up the Ariane 6 rocket programme, supporting smaller private launch startups like Isar Aerospace and Rocket Factory Augsburg, and building new launch infrastructure in French Guiana and northern Sweden. The goal is straightforward: stop depending on American rockets to get European satellites into orbit.

Samantha: The context here is uncomfortable. For the past two years, Europe has had to book launches with SpaceX for several flagship missions, including parts of the Galileo navigation system, because Ariane 6 ramped up slowly and the smaller Vega rocket had reliability problems. That's a sovereignty issue when those satellites carry military and critical infrastructure payloads. The new package is explicitly framed as catching up, not winning, but officials say Europe cannot afford to fall further behind.

Bob: A notable shift in this announcement is the embrace of competition. Instead of channelling everything through ArianeGroup, the traditional prime contractor, the funding deliberately backs multiple private players to drive down costs and speed up innovation. That's a model borrowed from how NASA approached SpaceX a decade ago. Whether European startups can deliver on that timeline is the open question, because building reliable rockets is famously hard and famously slow.

Samantha: For European industry, this matters because satellite launches underpin everything from broadband internet and earth observation to climate monitoring and defence. If Europe can launch reliably and affordably, it keeps that economic value at home. If not, the gap with the US and China widens further. The first tranche of the funding will flow in early 2027, with milestones tied to demonstrated launches rather than just promises.

Deepfake voice fraud hits European bank customers, and regulators are scrambling to respond.

Bob: European banking associations issued a coordinated warning yesterday about a sharp rise in deepfake voice fraud targeting retail customers. Criminals are using AI-generated audio that clones a family member's voice, often a child or grandchild, to call elderly account holders and convince them to authorise urgent transfers. Several national fraud reporting centres reported losses in the tens of millions of euros over the past quarter alone, with Germany, France, and the Netherlands hit hardest.

Samantha: The scam itself is an old trick with new technology. What changed is the quality and the accessibility of voice cloning tools. A few seconds of audio scraped from social media is now enough to generate a convincing imitation, and the calls are increasingly hard to distinguish from real ones, even for people who know the supposed caller well. Banks are responding with extra friction on urgent transfers, mandatory cooling-off periods, and prompts that ask customers to verify through a separate channel.

Bob: Regulators are now pushing for two things. First, faster intervention powers for banks to freeze suspicious transfers without waiting for customer confirmation, which raises its own questions about consumer rights. Second, stronger obligations on the platforms hosting voice-cloning tools to verify users and watermark outputs. The AI Act has provisions for synthetic media labelling, but enforcement on the consumer fraud angle has been patchy at best, and industry groups argue current rules lag well behind the criminal use cases.

Samantha: For listeners, the practical advice from fraud prevention agencies is simple: if you get an urgent call from a relative asking for money, hang up and call them back on the number you already have. A codeword agreed with family members in advance is another low-tech defence that consumer protection groups are actively promoting. The technology is moving faster than the rules, and for now, the burden is landing squarely on customers.

Opal is the focus app that uses AI to block your most distracting apps automatically.

Bob: To close out, two things you can actually try today. The first is Opal, a focus app for iPhone and Android that has been getting a lot of attention in productivity circles. The idea is to block distracting apps like Instagram, TikTok, and email during set work hours, but with a twist: it uses on-device AI to learn your patterns and suggest focus sessions based on when you actually get pulled away. Users report it cuts screen time by significant margins within the first week.

Samantha: What sets Opal apart from older screen-time tools is how it handles the willpower problem. Instead of just nagging you, it offers what it calls deep focus mode, which makes it deliberately annoying to unblock an app mid-session, requiring you to type out a reason or wait a cooling-off period. There's a free tier that covers the basics, and a paid version with more detailed analytics. It works in any European country and supports multiple languages, including French, German, Dutch, and Spanish.

Bob: One thing to know upfront: the deeper blocking features require giving the app screen-time permissions, which means it can see which apps you open and when. The company says all of that processing happens on your device and is not uploaded, but it's worth checking the privacy settings before you commit. Reviewers consistently praise it as one of the more effective focus tools currently available, particularly for people who already tried the built-in screen-time limits and found them too easy to swipe past.

Samantha: A good fit for anyone who lost an afternoon to doom-scrolling last week and wants a structural rather than willpower-based fix. Download is free, setup takes about five minutes, and you can run a first focus session straight away.

Darktable is the free open-source photo editor that rivals expensive professional software.

Bob: The second find is Darktable, a free and open-source photo editing application that just shipped a major update. It runs on Windows, Mac, and Linux, and it's aimed at people who shoot in raw format, so think hobbyist photographers, travel shooters, and anyone tired of paying a monthly subscription to Adobe. The latest version brings improved noise reduction, better colour grading tools, and a faster interface for large photo libraries.

Samantha: The appeal is straightforward: it does roughly what Adobe Lightroom does, but without the subscription and without sending your photos to a cloud service. Everything stays on your own machine, which matters for photographers who shoot sensitive material or just prefer to own their tools. The learning curve is steeper than Lightroom, and the interface can feel busy at first, but tutorial videos from the active community make the ramp-up manageable within a weekend.

Bob: Notable in this update is the addition of AI-assisted masking, which lets you select a sky, a person, or a specific colour range with one click and adjust just that area. That used to be a paid-software feature, and it now works locally without any subscription. For European photographers concerned about data residency and cloud lock-in, Darktable is one of the most credible free alternatives currently available, and the update is rolling out worldwide as of this week.

Samantha: Today we covered: the G7 debate on sharing advanced AI with trusted partners, a major European cloud outage exposing concentration risk, six billion euros for Europe's own rocket capacity, deepfake voice fraud hitting bank customers, and two things to try yourself: the focus app Opal and the open-source photo editor Darktable.

Bob: Want to know more or react? Visit stateoftech.eu or email us at info@doorzetters.net.

Bob: State of Tech, the tech world in 15 minutes.