The last full week of May 2026 was a lot, even by 2026 standards. A near-trillion-dollar funding round. A papal encyclical. A fresh pile of security holes. A buildout running headfirst into the power grid. Here are the ten stories that mattered most, ranked by how much attention each drew across the major outlets, counting up to the biggest.
10. The backlash against AI at work got louder
MIT Technology Review pushed back on the “AI jobs hysteria” while flagging a quieter erosion of entry-level roles, and the cultural mood turned sharper than the raw numbers suggest. US law enforcement started flagging “anti-tech extremism” as a threat category, and DuckDuckGo said installs jumped 30 percent after Google’s AI search redesign. People are not just worried about AI now, some are actively opting out of it.
9. YouTube and Google moved to label AI content automatically
YouTube will now detect and label photorealistic AI video on its own rather than trusting creators to disclose it, and it is making those badges harder to miss on Shorts and long-form clips. Google pushed its SynthID watermark further with a new content-detection API, with Nvidia and OpenAI already signed on. Provenance is quietly turning into infrastructure instead of an afterthought.
8. Humanoid robots crossed from demo reels into real factories
Figure announced a commercial deployment and said it is building humanoids at unprecedented scale, while China rolled out digital ID numbers to track humanoid units across their lifecycle. The data pipeline industrialized alongside the hardware, with startups paying workers to wear sensors and record everyday tasks, and Nvidia open-sourced Cosmos 3 for physical-AI reasoning. The robots are no longer the hard part. Teaching them is.
7. US states and the EU tightened the screws on AI
Illinois passed what is being called America’s strongest AI safety law, a move that further weakens Washington’s grip on AI policy and hands other states a template to copy. In Europe, researchers found that every major chatbot failed EU compliance tests, a preview of tougher enforcement ahead. The regulatory vacuum is closing from the edges in.
6. The cost of AI became the story
Simon Willison summed up the month bluntly: AI got expensive, with usage now metered by the token and some users openly weighing whether to cancel subscriptions. The counter-move is efficiency, and it is moving fast: VentureBeat reported DeepSeek’s architecture is shredding the “token moat”, and researchers automated reasoning design to cut token use by nearly 70 percent. The race is no longer only about capability, it is about who can make intelligence cheap.
5. The buildout slammed into power, water, and money
Regulators told Europe to cool its data-center boom before water and electricity run short, even as SoftBank committed up to 75 billion euros to French data centers and Snowflake earmarked 6 billion dollars for AI chips. The constraint has clearly shifted from algorithms to infrastructure: compute, power, and cooling. Engineers are now optimizing their bills as aggressively as they optimize their models.
4. Agentic AI hit a hard production reality check
Gartner warned that most generative and custom-model projects will end up a bust, and analysts expect roughly four in ten enterprise AI agents to be demoted or scrapped. A new benchmark found frontier models scoring below 50 percent on real enterprise IT tasks, which is a polite way of saying a demo and a production system are different sports. The lesson is finally landing: build agents where they clearly pay off, not because a vendor pitched you eighteen of them.
3. The Pope wrote an encyclical about AI
Pope Leo XIV issued Magnifica Humanitas, the first papal encyclical dedicated to artificial intelligence, warning that the technology risks concentrating power in a handful of firms and, citing Gandalf, calling on humanity to “disarm” AI. The Vatican invited Anthropic to the rollout, which several observers read as a startling alignment between the Church and Silicon Valley, even as critics called it the most effective vendor lobbying in history. When the Vatican weighs in on your industry, the stakes have changed.
2. AI’s security problem stopped being theoretical
A critical flaw in Starlette left millions of AI agents exposed through a host-header auth bypass, Microsoft’s Copilot Cowork was caught exfiltrating files, and the curl project said AI-generated vulnerability reports now arrive more than once a day. On the offensive side, analysts warned that rogue states are already putting AI agents to work on sanctions evasion at industrial scale. As agents read and run untrusted code, prompt injection and supply-chain tricks have become the main event.
1. Anthropic raised 65 billion dollars and shipped Claude Opus 4.8
Anthropic closed a 65 billion dollar Series H at roughly a 965 billion dollar valuation ahead of a possible IPO, with run-rate revenue reported near 47 billion dollars per the company’s own announcement. It paired the raise with Claude Opus 4.8, an upgrade reviewers call modest but real, adding a cheaper fast mode and a new dynamic-workflow tool. The mix of mega-funding and a steady product cadence made it the defining story of the week.
That was the last week of May 2026: the money, the machines, and the morality all moving at once. See you next week.