Tuesday, May 26 May 26, 2026

Good morning. It’s Tuesday, May 26th.

Today’s AI story is a reality check.

Over the last day, the biggest signal is not another flashy model demo. It’s that institutions across business, labor, security, and governance are all trying to answer the same question: now that AI is moving from experiment to operating layer, who captures the value, who absorbs the risk, and who sets the rules?

First, Uber is saying the quiet part out loud. In a fresh Verge report, Uber president and COO Andrew Macdonald says AI spending is getting harder to justify because the company still can’t clearly tie soaring Claude Code token consumption to a measurable increase in useful consumer features. That matters because it cuts against the default assumption that more AI usage automatically means more output. Uber is effectively saying: impressive internal usage charts are not the same thing as shipped customer value. For enterprise buyers, that’s a strong reminder to demand ROI discipline, not just AI adoption theater.

Second, TechCrunch reports that ClickUp has laid off 22 percent of its workforce while framing the move as an AI-first organizational redesign rather than classic cost cutting. CEO Zeb Evans says the company is introducing thousands of internal AI agents, shifting employees toward directing and reviewing agent output, and even opening the door to million-dollar pay bands for workers who create outsized impact with AI. The core message is stark: companies are beginning to redesign org charts around agents, not just add copilots to existing jobs. But the same report notes Gartner data suggesting many companies cutting jobs around autonomous tech are not yet seeing clear financial returns. So the labor model is changing faster than the proof.

Third, the governance conversation jumped a level. Pope Leo XIV released his first major papal document, Magnifica Humanitas, calling for legal and ethical frameworks to govern AI. The Verge says the encyclical focuses on labor disruption, AI-powered warfare, children’s exposure to AI systems, and the need to keep human dignity at the center of decision-making. Anthropic also published remarks from co-founder Chris Olah, who spoke at the Vatican presentation and argued that AI questions are too big to leave to computer scientists alone. Whether or not you follow church politics, this is notable because AI governance is no longer just a fight among labs, regulators, and think tanks. It is now being framed as a civilizational and moral issue by one of the world’s largest institutions.

Fourth, the legal system is starting to feel the operational side effects. In a fresh Verge item, Robert Hart points to AI-assisted lawsuits flooding court dockets. The upside is greater access for people who cannot afford legal help. The downside is that already strained courts are now dealing with a growing wave of low-cost, AI-amplified filings. This is one of the clearest examples of what happens when generative AI lowers the cost of participation in a system that was never designed for near-zero-cost volume.

Fifth, security remains the most underappreciated constraint. In TechCrunch, Google Cloud COO Francis de Souza argues that there is no AI strategy without a data strategy and a security strategy, and warns that shadow AI, model sprawl, agent access, prompts, and old forgotten repositories all expand the attack surface. One especially sharp point: internal agents can surface neglected data stores that humans had effectively forgotten existed. That means the risk is not just model misuse. It’s agent-driven discovery of weakly governed systems inside the enterprise itself.

And related to that, The Verge highlighted a newer adversarial pattern: attackers exploiting chatbot personalities rather than just trying simple prompt jailbreaks. In other words, the social engineering layer is moving inside the model interface. That is a useful reminder that safety is no longer just about hardening a model at the system prompt level. It’s about defending the entire interaction style users develop with these systems.

So the through-line this morning is simple: AI is leaving the demo phase. Boards want ROI. CEOs want leverage. Institutions want guardrails. Courts want capacity. Security teams want control. And workers want to know whether AI is becoming a tool, a manager, or a replacement.

One business idea to watch: an AI governance and ROI audit service for mid-market companies rolling out coding agents and internal copilots. The product would combine usage telemetry, cost-per-workflow analysis, shadow-AI discovery, and lightweight policy controls, then turn that into a board-ready scorecard showing where AI is actually producing value and where it is quietly increasing legal, labor, or security risk. The buyers are CFOs, CIOs, and private-equity-backed operating teams. Why now: Uber’s skepticism, ClickUp’s restructuring, and Google’s security warnings all point to the same budget moment. What makes it defensible is workflow-specific benchmark data and a repeatable audit layer that sits above whichever models the client uses.

That’s your AI morning briefing for May 26th.