AI-generated CSAM surge in 2025 & Chip stocks and AI spending boom - Tech News (Apr 27, 2026)

We’ll start with a grim but important story: reports of realistic AI-generated child sexual abuse material are rising sharply. The Internet Watch Foundation says it received far more reports in 2025 than the year before, and it’s not just images—synthetic video is exploding. What’s especially alarming is where this content is surfacing, including AI companion sites and even advertising placements on mainstream social networks. The UK’s Online Safety Act puts removal obligations on platforms, but the IWF argues there’s still a major gap: companies aren’t required to run and share meaningful AI safety testing before deploying tools at scale. The government says it plans to go further by criminalising AI tools designed to generate CSAM, and even “how-to” manuals that teach people to abuse children using AI.

On the markets side, the AI trade has found its swagger again. Strategists on Wall Street are pointing to soaring chip stocks and record highs for major indexes as evidence that investors still believe the AI buildout has a long runway. Nvidia briefly touched a staggering valuation milestone, while Intel notched its biggest one-day pop in decades—both signs of intense demand for the hardware that keeps AI running. The big question isn’t whether spending is happening—it’s when it slows. Analysts say hyperscalers are planning enormous outlays on data centers, compute, memory, and networking, and that’s blurring the old boom-and-bust cycles the semiconductor world is known for. If you’re looking for the subplot, keep an eye on the less-glamorous winners too: power and grid infrastructure companies that enable the data-center surge.

That same compute arms race is showing up in the funding frenzy around Anthropic. Reports say Google is lining up a massive new investment, following Amazon’s recent commitment, with additional funding tied to performance milestones. The headline isn’t just the eye-watering valuation—it’s the strategic loop underneath: big cloud companies fund top AI labs, and those labs spend heavily on the backers’ cloud platforms to train and run models. Anthropic’s rapid growth has also come with growing pains, including strain on capacity that has reportedly contributed to outages and tighter usage controls. The takeaway is simple: AI demand is rising faster than the industry’s ability to deliver stable, always-available service—and that gap is now a board-level problem.

Now to geopolitics, where China is tightening the screws on cross-border AI deals. Beijing has reportedly blocked Meta’s planned acquisition of Manus, an AI startup associated with autonomous “AI agents.” Officials ordered the deal withdrawn, and the message is hard to miss: China-linked AI assets are increasingly treated as strategic infrastructure, not just venture-backed software. For Meta, it’s a setback in the race to secure talent and technology for the next wave of AI products. For the broader market, it’s another reminder that M&A in frontier AI is going to be shaped as much by national security logic as by business logic—especially when money, models, and data cross borders.

Inside companies, the AI shift is also changing how work gets measured—and not everyone loves it. A growing trend dubbed “tokenmaxxing” describes firms encouraging heavy use of AI coding tools, and in some cases tracking usage internally. The risk is obvious: when a metric becomes a target, people start optimizing for the number, not the outcome. Still, leadership at some companies argues the early stage is exactly when you want lots of real-world usage, because it reveals where AI actually helps—and it can even feed training data for in-house models. In that same spirit, Shopify is expanding its internship program dramatically, betting that early-career engineers who grew up building with AI will be more valuable, not less. The cultural tension is real, though, and it shows up in personal stories too—like a recent account from a designer-engineer who quit a well-paid job, describing AI-normalized workflows as alienating when nobody verifies outputs and consent gets fuzzy around things like automatic meeting transcription.

Palantir is dealing with a different kind of internal tension: employees and alumni are increasingly questioning whether the company’s government work is enabling abuse—especially around immigration enforcement. Reports describe staff asking for clearer safeguards and transparency about how Palantir’s software is used by agencies like DHS and ICE, and also expressing alarm over military applications tied to civilian harm. One detail that stood out: employees say internal debate has felt constrained at times, including messages disappearing after a short retention window in a key Slack channel. Palantir leadership has defended its role through internal sessions, but even internal privacy staff reportedly acknowledged a hard truth of modern enterprise software: a determined, malicious customer can be difficult to stop in advance, leaving audits and accountability after the fact as the primary checks. That’s a tough answer for employees who want stronger guarantees up front.

In Ukraine, we’re seeing a clear glimpse of where warfare is heading. Ukrainian forces are expanding the use of remote-controlled unmanned ground vehicles—robots that can haul supplies, evacuate wounded personnel, and even take part in combat tasks—alongside aerial drones. Leaders have pointed to operations where territory was regained with no Ukrainian infantry losses, and commanders say the goal is to move some of the most dangerous frontline work from humans to machines. It’s also fueling a fast-growing domestic robotics industry built under battlefield pressure. The strategic benefit is obvious—fewer casualties. The concern, raised by experts, is that more distance can lower the psychological and political threshold for using lethal force, and can raise the stakes for civilian safety if targeting or oversight fails.

A policy story to watch in North America: Manitoba’s premier says the province will introduce legislation aiming to ban youth from using social media platforms—and explicitly, AI chatbots too. Details like the minimum age and enforcement approach aren’t set yet, but the proposal reflects a growing willingness by governments to treat algorithmic feeds and conversational AI as public-health issues for kids, not just parenting challenges. Canada would not be alone here; other jurisdictions have moved toward age-based restrictions, and the debate is shifting from “should we” to “how could we possibly enforce it,” especially when teens can quickly migrate to less regulated apps and services.

A more hopeful milestone comes from biotech. Intellia says its one-time CRISPR-based therapy for hereditary angioedema hit its main goal in a pivotal Phase 3 trial, cutting attack rates dramatically and leaving many patients attack-free months later. If regulators agree with the safety and efficacy profile, this could become one of the most important proof points yet for in vivo gene editing—where the edit happens inside the body, rather than modifying cells outside the body and reinfusing them. It’s also a reminder that the biggest technology leaps aren’t always in apps or devices; sometimes they’re in medicine, where a single treatment could replace years of chronic management.

Finally, a fast-moving energy story with big tech implications. Europe’s energy security has been rattled by the disruption around the Strait of Hormuz, which cut flows of Persian Gulf oil and LNG. The consequence is a sharper push to diversify—more wind and solar in the grid mix, continued reliance on nuclear, and growing interest in newer nuclear designs and green hydrogen as strategic buffers. For the tech world, this matters because data centers and electrification targets don’t exist in a vacuum. If energy supply is volatile, AI infrastructure planning gets harder—and the politics around power generation get even more consequential.