WiFi Identifies People Through Walls: EU Privacy Alert

Samantha: Welcome to The State of Tech, The European Edition, Monday May twenty-five, 2026. I'm Samantha Lawrence.

Bob: And I'm Bob Russell. Today: Telegram lets AI bots run your inbox, Pope Leo the Fourteenth publishes his first AI encyclical, Anthropic closes a record-breaking funding round, the UK and Australia sign an AI security pact, ordinary WiFi turns out to recognise people with near-perfect accuracy, and a clever messaging trick you can try this afternoon. Let's start with Telegram.

Telegram rolls out AI assistant bots that can read, filter and reply to your messages.

Samantha: Telegram has launched a new generation of AI assistant bots that go far beyond the chatbots we're used to. These bots can read your incoming messages, filter what matters, and even reply on your behalf, based on permissions you set yourself. It's pitched as automated inbox management, but really it's an AI agent living inside your chat app.

Bob: And that's the shift here. Telegram isn't bolting AI on as a feature you open separately. It's embedding AI as a layer underneath every conversation. So instead of you scrolling through hundreds of messages, the assistant flags what's urgent, handles the routine replies, and lets you focus on the threads that actually need a human.

Samantha: The permission piece is the interesting bit. You decide what the bot can see, which chats it can touch, and how far it can go in replying. So in theory, a work group can stay off-limits while the assistant cleans up your newsletter clutter or confirms appointments.

Bob: In practice though, this raises real questions. A bot that reads your messages is, by definition, sitting on top of your private communications. Telegram has always sold itself on privacy, and now it's asking users to trust an AI layer with the most sensitive content on their phone.

Samantha: This is also a competitive move. WhatsApp, Signal, iMessage, they're all circling the same idea. Whoever gets messaging-plus-AI right first sets the standard for how a billion people handle their inbox every day.

Bob: And the bottom line is, we're watching messaging apps turn into operating systems for personal productivity. Your chats become your to-do list, your calendar, your inbox triage, all in one place.

Samantha: For European users, this lands in interesting territory. Telegram is hugely popular across Eastern Europe and growing fast in Western markets. An AI that reads private messages will draw attention from data protection authorities, especially around consent and what happens to message content once the bot has processed it.

Bob: Expect questions from regulators about whether user-defined permissions are genuinely informed consent, or just another tick box. The EU AI Act puts obligations on systems that interact with users at scale, and a bot reading your chats absolutely qualifies. If you're going to try this, read the permissions screen carefully. That's not advice we usually give for a messaging update.

Pope Leo the Fourteenth publishes his first encyclical, dedicated entirely to artificial intelligence.

Samantha: Pope Leo the Fourteenth has unveiled Magnifica Humanitas, his first encyclical, and it's entirely about artificial intelligence. He presented it at the Vatican alongside Anthropic co-founder Christopher Olah, which is itself a striking image, the Pope sharing a stage with one of the architects of modern AI.

Bob: The document focuses on human dignity in an age shaped by AI. It addresses AI's impact on human identity, on labour, and crucially, on the concentration of power among a handful of tech giants. The Pope draws explicit parallels with the Industrial Revolution and argues we need robust ethical frameworks before, not after, the damage is done.

Samantha: The fact that the Pope chose AI as the subject of his inaugural encyclical is the headline. Encyclicals shape Catholic teaching for decades. By picking AI, Leo the Fourteenth is signalling that this is, in the Vatican's view, the defining moral issue of our time.

Bob: And inviting Christopher Olah to the presentation is a deliberate choice too. Olah is known for his work on AI interpretability, basically, trying to understand what's actually happening inside these models. The Vatican isn't dismissing AI, it's engaging with the people building it.

Samantha: The power concentration angle is the part regulators will notice. When the Pope says it's a problem that a few companies hold this much influence, that lands differently than when an academic says it.

Bob: Key takeaway here: this gives political cover to lawmakers who want to push harder on antitrust and AI governance. The moral argument is now formally on the table.

Samantha: Italy, Spain, Poland, Ireland, France, all countries with significant Catholic populations and active AI policy debates. Expect MEPs and national lawmakers to quote from this document in the months ahead.

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.

Anthropic closes a thirty-billion-dollar funding round at a valuation above nine hundred billion.

Bob: Anthropic is on track to close a funding round this week worth more than thirty billion dollars, at a pre-money valuation north of nine hundred billion. If it lands, that would make Anthropic the most valuable private AI company in the world, ahead of OpenAI.

Samantha: The round is being co-led by Sequoia, Dragoneer, Altimeter and Greenoaks, each putting in roughly two billion. Existing investors are also piling back in. The signal is clear, the smart money still believes the frontier AI race is wide open.

Bob: Nine hundred billion dollars for a private company is a number that would have sounded absurd two years ago. To put it in perspective, that's larger than the market value of most companies in the FTSE 100.

Samantha: And it's happening at a moment when there are still real questions about how these AI companies actually make money sustainably. Compute costs are enormous, the talent war is brutal, and the revenue picture, while growing, doesn't yet match these valuations.

Bob: What this funding buys is time and compute. Anthropic can now train bigger models, hire more researchers, and outlast competitors who can't raise on these terms. It widens the gap between the top three or four labs and everyone else.

Samantha: The bottom line is, the frontier AI race is becoming a contest between a handful of companies with effectively unlimited capital. That has implications for innovation, for competition, and for which countries get a seat at the table.

Bob: For Europe, this is uncomfortable reading. Mistral in France is the closest thing to a frontier lab on the continent, and it's nowhere near these numbers. The strategic conversation about European AI sovereignty just got harder.

Samantha: Expect renewed pressure on Brussels and on national governments to back European AI infrastructure. Whether that translates into actual capital at this scale is the open question.

The UK and Australia formalise a deeper alliance on frontier AI security risks.

Samantha: The UK and Australia have signed a Memorandum of Understanding on AI security. The UK AI Security Institute and the Australian AI Safety Institute will share information on frontier AI capabilities, collaborate on evaluation methods, and exchange research. Ministers met in Canberra to sign the pact.

Bob: The framing is pre-emptive. They want to get ahead of AI-powered cyber-attacks, while also using AI to strengthen their own defences. It's an explicit recognition that frontier AI is now a national security matter, not just a tech policy one.

Samantha: This builds on the network of AI Safety Institutes that emerged from the UK's Bletchley summit. Bilateral deals like this are how that loose network turns into actual operational cooperation.

Bob: And the choice of Australia is significant. Five Eyes intelligence sharing already exists between the UK, US, Canada, Australia and New Zealand. This extends that logic into AI safety, deliberately keeping the work inside trusted security partnerships.

Samantha: The practical effect is faster, more structured exchange of what these governments learn when they test frontier models. If one institute finds a dangerous capability, the other knows quickly.

Bob: Key takeaway: AI safety has graduated from conferences and white papers to government-to-government agreements with operational teeth.

Samantha: The EU AI Office in Brussels is the obvious European partner for this kind of cooperation. The question is whether Brussels can move fast enough to plug into these arrangements, or whether individual member states will strike their own deals.

Bob: France, Germany and the Netherlands all have growing AI safety expertise. Watch this space, because the architecture of global AI safety is being built right now, and Europe needs to decide where it stands.

German researchers show ordinary WiFi can identify individuals with near-perfect accuracy.

Bob: To close out, two things you can think about and try today. First, a research breakthrough from Germany that's genuinely unsettling. Scientists have demonstrated that ordinary WiFi signals can identify specific people in a room with near-perfect accuracy. No camera, no special hardware, just the WiFi router you already have.

Samantha: The technique analyses how radio waves bounce off bodies. Everyone disturbs the signal in a slightly unique way, based on shape, height, posture, gait. The researchers trained an AI to recognise those patterns, and it can tell people apart through walls.

Bob: What you can do today is straightforward, check what's actually broadcasting in your home. Log into your router, see which devices are connected, and think about which rooms have WiFi coverage you didn't realise extended that far. Most people have no idea their router signal reaches into the neighbour's flat.

Samantha: There are also router settings worth reviewing. Many modern routers have a feature called "WiFi sensing" or motion detection, often turned on by default. You can switch it off in the admin panel. It takes about five minutes if you know your router password.

Bob: The broader point is that the devices we treat as invisible infrastructure are becoming sensors. Your router, your smart speaker, your doorbell, they're all collecting more than you think.

Samantha: And right now, there's no clear legal framework covering WiFi-based identification. It falls between cracks in privacy law because no personal data, in the traditional sense, is being collected. The signal itself becomes the fingerprint.

Bob: So the practical action: open your router's admin page, usually by typing 192.168.1.1 into your browser, find the wireless or advanced settings, and look for anything labelled sensing, motion or presence detection. Switch it off if you don't use it. Costs nothing, works on virtually every modern router.

Samantha: And under GDPR, expect European regulators to start asking hard questions about this. The research is German, the data protection conversation will follow quickly.

A simple Telegram trick that turns the app's new AI bots into a personal news filter.

Samantha: And second, sticking with Telegram for a moment, because today's lead story actually gives you something to try this afternoon. The new AI assistant bots can be configured to do something genuinely useful, filter news channels.

Bob: Telegram is full of news channels, and most people subscribe to far too many. The new assistant bots can be told to scan a list of channels, pull out only stories matching keywords you care about, and deliver one summary message a day.

Samantha: Setting it up takes maybe ten minutes. You open the assistant bot, give it permission to read a specific folder of channels, set your keywords, and choose a delivery time. The bot then handles the rest in the background.

Bob: The clever bit is you can layer keywords. So "AI" plus "Europe" plus "regulation" gives you a much tighter feed than just "AI". You stop drowning in noise and start getting the three or four stories a day that actually matter to you.

Samantha: This is what the AI agent hype has been promising for two years, and it's quietly arriving inside an app a billion people already have on their phone.

Bob: It's also a preview of where this is going. Today it's news filtering. Next year it's calendar management, expense tracking, and probably booking your holiday.

Samantha: It's free, it works on iOS and Android, and it's live in all European markets from today. Just remember the permissions point we made earlier, only give the bot access to channels you're comfortable with it reading.

Bob: And if you try it, give yourself a week before you judge it. These bots learn from how you interact with the summaries, so the first day's output is rarely the best.

Samantha: Today we covered: Telegram's new AI inbox bots, the Pope's first encyclical on AI, Anthropic's record valuation, the UK-Australia AI security pact, WiFi as a surveillance tool, and a Telegram trick for filtering your news feed.

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.