Big tech bets billions on AI biology, Claude Mythos finds zero-days, Tim Cook exits Apple

Samantha: Welcome to The State of Tech, The European Edition, Tuesday April twenty-one, 2026. I'm Samantha Lawrence.

Bob: And I'm Bob Russell. Today: big tech is pouring billions into the convergence of AI and biology, Anthropic's new Claude Mythos model has US bank bosses genuinely rattled, Tim Cook is stepping down at Apple after fifteen years, emerging tech is reshaping global warfare according to Malaysia's defence minister, Chinese streamer iQIYI faces a revolt over its AI actor database, and to close, KAIST researchers say they've finally cracked the crease on foldable phones. Let's start with the story that's moving the most money and the most labs this week.

Big tech just went all-in on biology, and the numbers are staggering.

Samantha: The lead story today comes from an ARK Invest analysis by research analyst Shea Wihlborg, and it pulls together four announcements from just the past week. The thesis is simple: AI's biggest impact may not be chatbots or coding assistants. It'll be biology, medicine, and healthcare. And the scale explains why. The human body has about thirty-five trillion cells, each with a genome of roughly three point two billion base pairs. The space of potential drug-like molecules is estimated at more than ten to the sixtieth power. Clinical trials fail around ninety percent of the time. Drug development takes a decade or more.

Bob: And here's what happened this week. AWS launched Amazon Bio Discovery, with more than forty biology models and agentic assistants that route AI-designed candidates to lab partners like Twist Bioscience. In a project with Memorial Sloan Kettering, they designed about three hundred thousand novel antibody candidates, narrowed them to one hundred thousand, and sent them to the lab. Something that used to take a year, done in weeks.

Samantha: Then OpenAI announced a strategic partnership with Novo Nordisk covering drug discovery, manufacturing, and supply chain. They also unveiled GPT-Rosalind, a frontier reasoning model built for biology and translational medicine, with Amgen and Moderna as early access customers. Anthropic, meanwhile, appointed Novartis CEO Vas Narasimhan to its board, the first pharma executive to join. And they reportedly acquired computational biology startup Coefficient Bio for around four hundred million dollars.

Bob: And Nvidia and Eli Lilly launched an AI co-innovation lab back in January, up to one billion dollars over five years, running on BioNeMo and more than a thousand Blackwell Ultra GPUs. Four different plays: AWS is the orchestration layer, OpenAI is doing enterprise pharma deals, Anthropic is buying biology talent, Nvidia is selling the picks and shovels.

Samantha: The bottom line is that drug discovery has gone from a specialist niche to a strategic battleground for every major AI player. If ninety percent trial failure drops to seventy, that's a generational shift in healthcare economics.

Bob: And look at who's at the centre of this. Novo Nordisk is Danish, Novartis is Swiss. Europe's biggest pharma champions are plugging directly into American AI infrastructure. That raises real questions about where the compute lives, where the data sits, and whether Europe ends up a buyer rather than a builder in the next wave of drug discovery. The EU's AI Act will have to grapple with scientific models that design molecules autonomously, and regulators at the EMA need to catch up fast.

An AI that finds decades-old vulnerabilities nobody else spotted.

Bob: Next up, Anthropic's new model, Claude Mythos, has triggered a genuine alarm in Washington. The preview release demonstrated advanced reasoning that autonomously identified high-severity software vulnerabilities, and wrote working exploits, for flaws that had been sitting in operating systems and web browsers for decades. Traditional tools missed them.

Samantha: US Treasury Secretary Scott Bessent and Federal Reserve Chairman Jerome Powell convened a surprise meeting with major bank CEOs. That alone tells you how seriously this is being taken. The concern isn't theoretical. If a frontier model can find zero-days at scale, the offensive-defensive balance in cybersecurity tilts in ways regulators haven't modelled. Banks are sitting on legacy software stacks that would take years to fully audit.

Bob: And the awkward part is that Anthropic positions itself as the safety-focused lab. This is exactly the kind of dual-use capability their Responsible Scaling Policy is meant to flag. The fact that it shipped, even in preview, means the industry's internal guardrails are being tested in public.

Samantha: Key takeaway: cybersecurity risk is no longer just an IT department problem. When the Treasury Secretary sits down with bank CEOs, it's a systemic risk conversation. And every critical infrastructure operator, from energy grids to hospitals, has to assume adversaries will get hold of similar capabilities.

Bob: European banks run on the same software stacks. The European Central Bank and ENISA will be watching this closely. Under DORA, the Digital Operational Resilience Act, European financial institutions already have strict requirements around ICT risk. Expect new guidance specifically on frontier AI within months. And GDPR adds another wrinkle: if an AI finds a vulnerability in a system holding EU citizen data, disclosure obligations kick in fast.

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.

After fifteen years and four trillion dollars, Tim Cook is handing over the keys.

Samantha: Tim Cook is stepping down as Apple CEO on September first. His successor is John Ternus, the fifty-year-old head of hardware engineering, who's been at Apple for twenty-five years and has led hardware since 2021. Cook becomes executive chairman. Under his tenure, Apple went from a successful phone maker to a four trillion dollar company and the most valuable brand on Earth.

Bob: And that's the tension right there. Ternus is a product and hardware veteran, not a software or AI specialist. Apple is widely seen as trailing on generative AI. Apple Intelligence has underwhelmed, the Siri overhaul has been delayed repeatedly, and rivals from OpenAI to Google are redefining what a smart device even means.

Samantha: Picking a hardware leader signals Apple still believes the device is the moat. The idea being that if you control the silicon, the sensors, and the operating system, you can eventually win on AI by integration, even if you're late to the model itself. That's a bet. Microsoft and Google are betting the opposite way, that the model is the product.

Bob: And Cook leaves a complicated legacy. Record revenues, yes, but also a company that's become deeply dependent on China for manufacturing and on the App Store for services revenue, both of which are under pressure.

Samantha: The bottom line is that Apple's next five years will be defined by whether Ternus can either build credible in-house AI or cut the right partnership deals. Rumours about a deeper tie-up with Anthropic or OpenAI will only intensify.

Bob: Europe is Apple's second-biggest market, and the relationship has been bumpy. The Digital Markets Act has forced Apple to open up the App Store, allow sideloading, and unbundle services. Ternus inherits that fight. And if Apple does partner with an American AI lab to power Siri, that means more EU citizen data flowing through third-party models. Expect the European Data Protection Board to pay very close attention.

Malaysia's defence minister says the digital backbone now beats the tank.

Bob: Story four. At the Putrajaya Forum, Malaysian Defence Minister Datuk Seri Mohamed Khaled Nordin laid out a blunt thesis: emerging technologies, AI, autonomous systems, quantum, space tech, and hypersonics, are fundamentally reshaping military power. Hardware, he said, is becoming secondary to the digital backbone, the compute, the data, and the AI that sits on top.

Samantha: His argument is that future battlefields are defined by connectivity, lethality, and precision, and the winner is whoever processes and acts on data faster than the adversary. This isn't a theoretical claim anymore. Ukraine has shown it in practice, with drone swarms coordinated by software, AI-assisted targeting, and commercial satellite intelligence folded into military decision loops in near real time.

Bob: And Malaysia, sitting between the South China Sea tensions and major shipping lanes, has every reason to say this out loud. What's notable is a smaller defence ministry explicitly framing AI and autonomy as the core of deterrence, not an add-on. That's a signal to regional partners and to arms suppliers about what they intend to buy.

Samantha: The implication for the global arms industry is huge. Traditional primes built their business on airframes, hulls, and armour. The growth area now is software, sensors, and AI integration. Companies that can't deliver a digital backbone will lose contracts.

Bob: And Europe is already moving. The European Defence Fund is channelling money into AI, autonomy, and quantum. Germany's Helsing, France's Mistral partnering with Thales, startups like Tekever in Portugal, they're all riding this wave. But Europe still lags the US and China on compute infrastructure for defence AI. And the ethical question, autonomous weapons, is sharper in Europe than anywhere. The EU will have to reconcile its values with the operational reality that adversaries won't wait.

Over a hundred Chinese celebrities just revolted against an AI casting database.

Samantha: Story five. iQIYI, one of China's biggest streaming platforms, debuted an AI artist database that lets producers use actors' likenesses in AI-generated dramas and films. The backlash was immediate. More than a hundred celebrities and huge numbers of fans took to social media to protest. Actors fear job displacement and the unauthorised reuse of their digital likenesses, even though iQIYI claims performers retain control.

Bob: And China is moving faster on AI-generated entertainment than anywhere else. State-backed studios have already released fully AI-produced short films. iQIYI's database is the logical next step, a structured marketplace of digital faces. The promise to actors is royalties without having to show up on set. The threat is that studios license a likeness once and use it forever, cutting human performers out of the loop.

Samantha: It's the same fight the Screen Actors Guild waged in Hollywood two years ago, playing out in a market with very different labour protections. Chinese actors don't have an equivalent union with real bargaining power. So this public backlash on Weibo is effectively their only leverage.

Bob: The bigger point is that every creative industry, film, music, publishing, games, is now negotiating the same question. Who owns a face, a voice, a performance style, once an AI can replicate it? And what does consent mean when it's buried in a contract clause?

Samantha: European performers' unions, from Equity in the UK to FIA across the continent, are already pushing for explicit consent, compensation, and transparency rules. The EU AI Act requires disclosure of AI-generated content, but doesn't yet fully address likeness rights. Expect France and Germany, with their strong cultural policy traditions, to lead on national rules. And European producers watching iQIYI will take notes on what not to do.

The crease on your foldable phone may finally be history.

Bob: And to close on something a bit lighter. Researchers at KAIST, the Korea Advanced Institute of Science and Technology, say they've developed a zero-crease technology for foldable displays. The visible, annoying line down the middle of every foldable phone since Samsung's first Galaxy Fold, apparently gone.

Samantha: KAIST has filed patents globally, including in the European Union. The team describes the solution as intuitive and simple, and crucially, compatible with existing manufacturing processes. That last bit is what makes this interesting. Plenty of lab breakthroughs die because they require an entirely new production line. If this slots into current factories, it could be in consumer devices within a year or two.

Bob: And the foldable market really needs this. After early excitement, sales have plateaued. The crease, the durability worries, and the premium price have kept foldables as a niche. Samsung, Huawei, Honor, and a handful of others have been iterating on hinges and polymer layers, but the crease has stubbornly remained. A clean fold changes the value proposition.

Samantha: The interesting part is that KAIST says it extends beyond phones, to foldable laptops and tablets. Which is where this gets genuinely useful. Nobody really needs a folding phone. A folding laptop that opens into a tablet, that's a different conversation.

Bob: And the bigger point is that hardware design still matters, even in an AI-obsessed moment. The device you hold in your hand shapes how you use software. A truly creaseless foldable could spark a new wave of product experimentation.

Samantha: With patents filed at the European Patent Office, licensing deals with European manufacturers are on the table. Think of laptop makers like Lenovo's European operations, or specialist designers in the Nordics and Germany. And European consumers have been cautious about foldables so far, exactly because of durability concerns. A zero-crease promise might finally move the needle.

Bob: Honestly, if I can finally stop squinting at that line, I'm sold.

Samantha: Today we covered: big tech's push into AI and biology, Anthropic's Claude Mythos rattling US banks, Tim Cook handing over Apple to John Ternus, Malaysia's defence minister on AI warfare, China's iQIYI and the AI actor backlash, and KAIST's zero-crease foldable breakthrough.

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.