The Pope’s AI Guy Is Worried About ‘Cartel’ Behavior Among Big Labs

Paolo Benanti, the Vatican's go-to advisor on AI ethics, just called out something most of the tech world is too polite to say out loud. According to WIRED, he thinks the existential risk crowd is missing the forest for the trees, and meanwhile, the big AI labs are starting to look a lot like a cartel.

The timing matters. While OpenAI, Anthropic, and Google trade white papers about hypothetical superintelligence risks, they're also quietly lobbying for regulations that happen to favor large, well-funded players. Benanti's argument is that obsessing over whether AI will destroy humanity in 2045 conveniently distracts from harder questions about who controls the technology right now, in 2026.

This is the cartel behavior he's flagging. A handful of companies control the compute, the data pipelines, and increasingly, the regulatory conversation. They set the terms for API access, decide what safety benchmarks matter, and shape public policy through sheer market presence. It's not a conspiracy, it's just how concentrated markets work when barriers to entry are measured in hundreds of millions of dollars.

Benanti's day job is advising the Pope on technology and human dignity, but his broader point applies to anyone building with AI tools. The governance gap isn't about sci-fi scenarios. It's about who gets to build, who gets to use the outputs, and whether we're creating public accountability or just letting a few companies write their own rules.

The public debate he's calling for isn't happening in most countries. We get product launches and safety theater, but not much discussion about whether frontier AI should be treated as critical infrastructure, whether model weights should be open, or how to prevent lock-in when one company's API becomes the foundation for thousands of businesses.

What this means for you: If you're building anything serious on top of frontier models, you're making a bet on whoever controls that API. Diversify where you can. A simple workflow: pick one key feature in your product and prototype it with two different providers (OpenAI and Anthropic, or Claude and Gemini). Then ask your AI assistant: "What would break in my product if [provider] changed their API terms or raised prices 3x? Give me a migration checklist." That exercise will show you exactly how much leverage you've handed over, and whether Benanti's cartel warning applies to your stack.