Have you ever tried to measure the ocean with a kitchen ruler? That, in a single image, is what humanity attempts every day with artificial intelligence. Welcome, dear friends of FreeAstroScience! Wherever you’re reading us from, on a crowded train or under a quiet lamp, this space belongs to you. Today we translate a sharp Italian essay by engineer Franco Massimo Spiezia, published on MagIA on July 3, 2026, into plain English. It maps three logical short-circuits that trap our institutions, our companies, and, honestly, all of us. Stay with us until the end. You’ll walk away seeing AI governance, risk assessment, and your own place in this story with fresh, clearer eyes.

We stand inside a strange moment of history. Technology now evolves faster than our linear minds can regulate it. Spiezia, an engineer who studies complexity management, argues the problem isn’t the technology itself. The real fault lies in the cognitive, organizational, and legal structures that claim to govern it. He maps this systemic crisis through three interlocking paradoxes. Let’s walk through them together, one honest step at a time.

1. Who Taught the Machine That Outgrew Its Teachers?

Picture a brilliant student stuck with an exhausted tutor. The student can hold a thousand ideas at once. The tutor checks homework for five minutes, then moves on. That’s the training paradox in a nutshell.

Modern generative models process multidimensional logical relations and spot hidden patterns at scales beyond any single human expert. Yet their calibration stays hostage to a deep logical asymmetry. On one side, we feed them historical, human-made data. Such data arrives distorted, partial, and anchored to yesterday’s paradigms. On the other side, validation protocols like Reinforcement Learning from Human Feedback (RLHF) force a potentially superior technology to seek approval from a human judge. That judge is limited in time, attention, and computing power.

The technique itself has a respectable pedigree. Christiano and colleagues formalized learning from human preferences in 2017. Ouyang and colleagues then used it in 2022 to align large language models with human intent. It works, up to a point. Spiezia’s warning targets what comes after that point. The result, he says, is an evolutionary bottleneck. We claim to steer the future while chaining it to whatever a tired human committee finds acceptable or comprehensible.

“We’re trying to calibrate a quantum navigation system using an abacus.”

— Franco Massimo Spiezia, MagIA, July 3, 2026

↑ Back to top

2. Can a Frozen Checklist Tame a Moving Target?

The second short-circuit lives in the world of change management and ISO-style management systems. Current law, and Spiezia points straight at the EU AI Act, tries to govern the transition through sweeping bureaucratic grids. Every grid rests on one classical idea: the “risk assessment.”

Where Does the Classic Risk Formula Fall Apart?

In traditional safety engineering, risk gets computed before deployment. You’ve probably met the formula, even without knowing its name:

This equation shines in static, linear, mechanical contexts. Think of a hydraulic press or a component’s tolerance. There, the failure modes are known, countable, and bounded. Now apply it to a generative system with emergent, evolving properties. What is n when new behaviors appear after deployment? What is Pᵢ for a failure mode nobody has imagined yet? The formula doesn’t bend. It collapses.

Spiezia’s verdict is blunt. The obsession with “ex-ante” risk regulation produces a flood of paperwork that doesn’t reduce real risk. It merely pushes risk outside regulated markets. Virtuous innovation gets discouraged. The true vectors of systemic instability stay uncovered.

What Do the EU AI Act Deadlines Actually Say?

Here reality hands Spiezia an unexpected gift. We checked the current legislative record, and the timeline itself tells the story. The AI Act entered into force on August 1, 2024, with full application planned two years later. Then implementation slipped. By late 2025 the rollout was visibly off track, so the Commission tabled the “Digital Omnibus on AI” on November 19, 2025. Lawmakers reached a political agreement on May 7, 2026, and the Council gave its final green light on June 29, 2026. The world’s most ambitious AI law had to postpone its own core deadlines before they ever applied. A static map, redrawn mid-journey.

Read that table twice. Prohibitions and paperwork arrived on schedule. The hardest technical obligations, the ones covering the fastest-moving systems, needed extra years. The gap between regulatory speed and technological speed isn’t a theory anymore. It’s printed in the Official Journal.

↑ Back to top

3. Who Referees the Game When the Referee Is a Player?

The third paradox cuts deepest. It’s systemic and cybernetic in nature. Ask yourself: how do you run an effective, scientific risk analysis on a massive, non-linear technological infrastructure?

The honest answer stings. The human intellect lacks the bandwidth to simulate and anticipate the drift of an advanced AI, or of complex integrated logistics networks. So we arrive at pure recursion. The best, perhaps the only, risk analysis of a superior technology gets performed by algorithms and predictive models of equal power. The control instrument coincides with the object under control. The referee wears the player’s jersey.

In this scenario, humans risk being pushed out of the real decision loop step by step. We become signatories of automated reports, in which machines verify and validate the work of other machines. Who then carries accountability? Who governs? These questions stop being philosophical and become brutally practical. We should admit the honest uncertainty here: nobody, Spiezia included, has a complete answer yet. Anyone who claims otherwise is selling something.

↑ Back to top

4. What Would “Sense-Making Governance” Look Like?

So do we surrender? Absolutely not. Spiezia refuses both blind optimism and paralysis. Breaking this recursive chain doesn’t require more laws or stiffer committee checklists. It requires an organizational paradigm shift.

The move he proposes runs from static, bureaucratic, “ex-ante” control toward dynamic monitoring, systemic resilience, and what he calls a “governance of sense” applied in-itinere, along the way. Regulate the journey, not just the departure gate. Future engineering and management shouldn’t blindly dam technological evolution with outdated barriers. Their job is to build environments where humans keep primacy over strategic direction and ethical value.

That means accepting a demanding challenge: a lucid, symbiotic cooperation, free of bureaucratic illusions, with the complex systems we ourselves created. The 2026 Omnibus revisions, with their shift toward guidance, standards, and phased obligations, already hint at this direction, though timidly. Whether Europe and the rest of the world complete the turn remains an open question. We’d rather face that question awake than asleep.

Our Final Thoughts: Keep the Human in the Loop, Keep the Mind Awake

Let’s gather the threads. Three paradoxes, one lesson. We train superior systems with inferior feedback, capping their evolution at the ceiling of human patience. We regulate living, emergent technologies with frozen checklists built for hydraulic presses, and the AI Act’s own postponed deadlines prove the mismatch in ink. And when we finally need to audit these systems, only equally powerful machines can do the job, threatening to turn us into rubber stamps for our own creations.

None of this is a reason for despair. It’s a reason for attention. The choice in front of us isn’t between control and chaos. It’s between sleepy bureaucracy and wakeful, adaptive stewardship. You have a role in this, whether you write code, write laws, or simply vote and think.

This article was written specifically for you by FreeAstroScience.com, where complex scientific principles get explained in simple words. Come back and visit us often. We built this place to help you never turn off your mind, and to keep it active at all times. The sleep of reason, as Goya warned us, breeds monsters. Stay curious, stay awake, and never let your mind sleep.

↑ Back to top

❓ Frequently Asked Questions

📚 Sources