The Vatican’s AI Guru Hasn’t Even Tried It

Cardinal Michael Czerny, who leads the Vatican’s commission on artificial intelligence, recently admitted he hasn't actually used the tools he is helping to regulate. He cited a lack of time, opting instead to rely on academic briefings and philosophical frameworks. It’s a surreal admission for someone shaping the moral compass of global AI development.

For us building products in Flanders, this is a painful reminder that the loudest voices in the room are often the least acquainted with the actual stack. When regulators—whether in the Vatican or the European Commission—talk about 'human-centric AI' or 'existential risks,' they are often operating in a vacuum of theory. They don't know the friction of getting a RAG pipeline to stop hallucinating or the reality of managing API costs while trying to ship a feature that actually helps an SME client sell more inventory.

This gap matters because it dictates the tenor of future compliance. When policymakers don't use the tech, they tend to over-index on abstract 'guardrails' that don't account for how we actually build. They view AI as a monolithic entity rather than a collection of tools, libraries, and trade-offs. The result is often legislation that treats a simple intent-classifier the same way it treats a foundation model, creating unnecessary overhead for small studios like ours.

Don’t wait for the bureaucrats to understand your dev workflow before you start documenting your own safety practices. You don't need to be a theologian to understand your product's impact, but you do need to be able to explain it to someone who doesn't code.

One thing to do this quarter: Audit your current AI features and document their 'logic chain' in plain language—not for the user, but for a hypothetical auditor. If you can’t explain how your model reaches a decision in three sentences, you have a liability problem, not just a technical one.

Source: Politico EU