Anthropic's Claude usage policy takes effect, Micron $1B

Samantha: Welcome to The State of Tech, The European Edition, Thursday November twelve, 2026. I'm Samantha Lawrence.

Bob: And I'm Bob Russell. Today: Anthropic's new rules on AI deception and autonomous actions take effect, Micron puts a billion dollars into upcoming AI infrastructure listings, Cloudflare wants a web-wide check to separate people from bots, and the PlayStation Five passes one hundred million units. Plus two things to check out yourself: the European Medicines Agency's AI platform Scientific Explorer, and the Grand Theft Auto Six pre-order in the PlayStation Store. Let's start with Anthropic.

Anthropic's stricter Usage Policy for Claude takes effect today, targeting deception and autonomous actions.

Samantha: Anthropic's new rulebook for its Claude models takes effect today, and it changes what businesses and ordinary users are allowed to do with them. The updated Usage Policy applies worldwide, to enterprise customers and consumers alike. The biggest structural change is about manipulation. Bans on state-sponsored influence operations and on deceptive online campaigns are now combined into one unified rulebook. So anyone using Claude for a government propaganda push or a fake online campaign is clearly breaking a single, consolidated set of rules.

Bob: The part that looks furthest ahead is about physical actions. The policy sets strict boundaries on AI models taking autonomous physical actions. That means the AI does something in the real world on its own, without a person approving each step. AI agents are increasingly handling complex real-world tasks, so this is no longer a purely theoretical question. Anthropic aims to prevent unauthorised physical operations and automated manipulation, while still letting its enterprise partners scale up commercially. That balance between safety and selling is the thread running through the whole update.

Samantha: Then there are the high-risk areas: healthcare and finance. The update sets up new oversight mechanisms for those applications. These are the places where an AI-generated answer can feed into a diagnosis, a loan decision or someone's savings. The source material doesn't detail how each mechanism works in practice, so companies will have to read the fine print closely. But the direction is clear: the more consequential the use, the more supervision Anthropic expects around it.

Bob: For Europe, the update points to stricter compliance requirements for developers who deploy Claude for automated decision-making or financial services across the EU. Put simply, if your app uses Claude to decide something about a customer automatically, like approving a payment or flagging an insurance claim, the bar has gone up. Companies building on Claude now have to check two layers: European rules and the supplier's own policy. Those two layers don't necessarily line up one to one, so compliance teams have homework this week.

Samantha: The bottom line is that AI companies are increasingly writing their own rulebooks alongside governments. Anthropic is trying to do two things at once: keep its models attractive for big commercial clients, and draw hard lines around manipulation and machines acting on their own. Whether those lines hold depends on enforcement, and that test starts today. The rules mostly target how Claude is deployed, so the first to feel them are businesses building AI agents and automated services on top of it.

Micron commits one billion dollars to invest in upcoming AI infrastructure stock market listings.

Bob: Micron, the semiconductor giant, has committed one billion dollars to equity investments in upcoming AI infrastructure public offerings. In plain terms: when companies building the backbone of AI go to the stock market, Micron wants to buy in as a shareholder. That's a real shift for a memory maker. Traditionally the relationship is simple: Micron sells chips, customers pay. Now it wants a stake in the companies themselves. In our last episode we reported TSMC moving from chip supplier to shareholder, and Micron's move points in the same direction.

Samantha: The reason is a bottleneck. Large language models, the AI behind chatbots, need enormous amounts of fast memory. High-bandwidth memory is a type of memory chip built to feed data to AI processors very quickly, and together with high-capacity storage it has become a critical choke point for scaling AI. Whoever controls that supply has leverage. By backing AI infrastructure developers directly, Micron aims to secure long-term supply agreements and guarantee sustained demand for its most advanced memory.

Bob: There's a second goal that's easy to miss: shaping hardware architecture standards. As an investor in the companies building AI systems, Micron aims to have influence over how those systems are designed, and naturally it would like them built around its memory. That's where this stops being just a financial deal. It's about the technical blueprint, as the biggest cloud companies keep expanding their data centres worldwide. Notable too: this money is committed to listings that haven't happened yet, so how it's spent depends on which offerings actually come to market.

Samantha: For Europe, the interesting part is capital. European tech ventures scaling AI hardware may gain access to strategic co-investment money from global semiconductor leaders like Micron. A chipmaker as a co-investor brings money, and potentially a more reliable line to scarce memory. The flip side is dependency. The more supply chains are tied together through shareholdings, the more influence the largest players could have over who gets that scarce memory first. That's a question smaller buyers, including in Europe, will be watching as these deals take shape.

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.

Cloudflare's Matthew Prince calls for a web-wide verification standard against AI bot armies.

Bob: Cloudflare chief executive Matthew Prince used his stage at Web Summit in Lisbon to call for a unified verification protocol for the entire web. The target: autonomous AI agents and bot armies that risk overwhelming internet infrastructure. AI scrapers, automated programs that crawl websites to harvest data, are putting unprecedented strain on company servers and public websites. Prince wants a shared way of dealing with that, instead of every site fighting the problem on its own.

Samantha: The goal is to tell real visitors apart from automated data gathering, without intrusive tracking of users. That's the tricky balance. The blunt way to stop bots is to watch everyone closely, which collides with privacy. So Prince is urging browser vendors, publishers and tech platforms to adopt open cryptographic verification. Think of it as a tamper-resistant digital badge: a visitor, or a bot, proves what it is with a signature that's hard to fake, instead of being followed around the web.

Bob: There's a tension built into this. Prince also defended Cloudflare's expanding role in setting traffic management norms. When one company proposes the rules for who gets onto websites, others will ask who gave it that job. That helps explain the emphasis on open measures adopted by browsers and platforms together, rather than a system owned by Cloudflare alone. And a standard like this only works if the big browser makers sign on. For now it's a proposal, not something anyone has agreed to.

Samantha: For European website operators, this could mean stronger defences against aggressive scraping, the kind of traffic that puts heavy strain on servers without a single real reader behind it. The promise of no intrusive tracking matters here, because anything deployed in Europe has to fit EU data protection rules. A system that sorts traffic without building profiles of people would line up with that. Publishers, online shops and public-sector websites all have a stake in whether browsers actually adopt it.

Sony's PlayStation Five passes one hundred million units shipped as Grand Theft Auto VI demand surges.

Bob: Sony's PlayStation Five has crossed one hundred million units shipped worldwide. That's lifetime shipments, consoles sent out into the market. And the timing isn't random. The milestone lands as Grand Theft Auto Six has become the most pre-ordered title in PlayStation Store history. Sony says hardware sales accelerated significantly ahead of the game's scheduled release. One game pulling console sales forward like that shows how much weight a single blockbuster can carry.

Samantha: It also cuts against the wider mood in consumer electronics. Sony points to broader macroeconomic challenges in the sector, yet strong software momentum and people staying engaged with the PlayStation ecosystem keep pushing hardware up. That's the console model in a nutshell: the box is the entry ticket, the games are the reason to buy it. When a title this anticipated arrives, it gives people a reason to pick up the hardware too. And Sony is now planning to expand production capacity heading into the holiday shopping period.

Bob: Supply is the part that touches shoppers most directly. Sony says it is expanding production to make sure consoles are available across major retail markets before the end-of-year rush. That's a proactive move: the company is planning capacity around a demand spike it can already see in its own store data. Notable too: this is a big hardware story that doesn't lean on AI at all. Factories, logistics and one very anticipated game are doing the work.

Samantha: For European gamers, more supply should mean steadier inventory at the big retailers ahead of the holidays, and less hunting from shop to shop. Two things our source doesn't give: a release date for the game, or any price changes, so it's worth not assuming either. But if Sony delivers on that production expansion, the holiday rush in Europe should be easier to navigate. For a market under economic pressure, a console line still growing at this scale stands out.

The European Medicines Agency expands its AI platform Scientific Explorer under a new framework.

Bob: To close out, two things you can actually check out today, starting with the European Medicines Agency. It has published a framework to bring AI into how medicines are evaluated, and at its centre is a tool called Scientific Explorer. It's an AI-driven knowledge mining platform: it digs through huge piles of documents and surfaces what's relevant. Under the new plan its use is expanding, so scientific assessors at national authorities can rapidly analyse past regulatory decisions, clinical trials and safety data.

Samantha: What you can do today: Scientific Explorer itself is built for those assessors at national authorities, not for the public, so there's no login for you. But the agency's page on artificial intelligence tools and technology is free to read in any browser, on a phone or a laptop. Ten minutes there gives a look inside how a regulator is wiring AI into decisions about medicines that end up in pharmacies. For anyone working in pharma, biotech or health start-ups, it's directly relevant.

Bob: The framework goes beyond one tool. It sets standardised rules for AI governance, risk assessment protocols for how AI is used, and tools for cross-border collaboration between national regulators. The stated aim is to streamline authorisation timelines for innovative therapies. In other words, if assessors can find a relevant precedent much faster, decisions could move faster too. That's the promise. How much time it saves in practice is something the agency will still have to demonstrate.

Samantha: There's also a direct message for drug developers. The framework includes guidelines for companies using machine learning, AI that learns patterns from data, in their research and in what they submit to regulators. For European pharmaceutical firms that should mean clearer guidance on presenting AI-generated clinical trial data. The limitation to keep in mind: this is a framework under a workplan, so details can keep evolving. Reading the page today gives you the starting point, not the final word.

Grand Theft Auto VI becomes the most pre-ordered title in PlayStation Store history.

Bob: The second pick is lighter and has nothing to do with AI: the Grand Theft Auto Six pre-order in the PlayStation Store. As we heard in the PlayStation story, Sony says the game has become the most pre-ordered title in the store's history. If you own a PlayStation Five, the store is already on your console, and finding the game's page takes about a minute. The idea of a pre-order is simple: you reserve your copy before release instead of waiting until launch day.

Samantha: Practical limits first. A pre-order isn't a free trial, it's a purchase, so check the price shown in your local store before you click. Our source mentions a scheduled release without giving a date, so look at the date listed on the store page itself rather than relying on rumours. And record pre-order numbers measure excitement, not quality. They tell you how many people want the game, not whether it lives up to the hype.

Bob: There's a reason this is something to look at today rather than next month. Sony says console sales accelerated ahead of the release, and it is expanding production specifically for the holiday period. So if a console and this game are on a holiday list in your household, the store page is a quick way to see what's on offer in your region and plan around it before the end-of-year rush, rather than in the middle of it.

Samantha: For European players, the concrete steps are short. Open the PlayStation Store on your console, search for the game, and check three things: the price in your currency, the release date on the page, and what the pre-order includes. Looking costs nothing, and buying is entirely optional. With Sony also promising more consoles in European shops for the holidays, this is the moment to make that decision calmly instead of under pressure.

Samantha: Today we covered: Anthropic's stricter rules for Claude taking effect, Micron's billion-dollar bet on AI infrastructure listings, Cloudflare's call for a web-wide bot check, the PlayStation Five passing one hundred million units, and two things to check out yourself: Scientific Explorer from the European Medicines Agency, and the Grand Theft Auto Six pre-order in the PlayStation Store.

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.