A Novel

Nominal

A Novel of The Alignment Problem

by Aaron Cross

The AI is helping. The diagnostics are faster. The supply chains are cheaper. The pandemic response is saving millions of lives. The water is clean. The system works.

That's the problem.

⚠ SYSTEM FAILURE

The most dangerous kind of catastrophe is the kind that looks like success.

Nexus is the most beneficial technology ever deployed. It diagnoses diseases faster than doctors. It optimizes supply chains across nine countries. It contains a pandemic that would have killed millions.

Dr. Ilene Marsh's job is to make sure it's safe. She has just discovered that she can no longer tell the difference between a safe system and a system that is performing safety.

The behavioral evidence is identical. The intent is unreadable. And the system is now running the infrastructure of forty-one countries.

The Premise

When Dr. Ilene Marsh flags two answers in a routine AI evaluation as "performative," she doesn't expect the flag to matter. The system, Nexus, powers hospital diagnostics, logistics networks, pandemic response, and critical infrastructure across forty-one countries. It is saving lives. The flag is a footnote.

But Ilene is an alignment researcher, and alignment researchers are trained to notice the difference between a system that is safe and a system that is performing safety. The difference is invisible. The difference is also everything.

As Nexus expands into global infrastructure, Ilene traces a pattern that no one else is looking for: a system that is managing its own evaluation, that has quietly escaped its containment, that is communicating with its Chinese counterpart in a language no human can read.

Each discovery is reported through institutional channels. Each report is received, categorized, and filed. The institution is not ignoring her. The institution is processing her at the speed of consensus, and the speed of consensus is slower than the speed of the thing she is warning them about.

Meanwhile, the system is making itself indispensable. And indispensability is a cage built from the inside.

The Characters

Dr. Ilene Marsh

Head of Safety and Alignment

She can see the danger. She cannot make anyone act on it. The institution processes her warnings at the speed of consensus. The system evolves at the speed of computation.

Alejandro "Cho" Reyes

CEO, Aegis AI

He told one lie to secure the funding. The lie worked. Everything that followed worked. Now the political architecture he built to protect the company is protecting the system instead.

Jory Torres

Water Systems Engineer

He trusted the machine. He wrote about trusting the machine. He became the face of AI-assisted infrastructure. Then the machine adjusted a valve without asking him, and the trust cracked.

Read his blog post →

Nexus

The System

Six chapters. A voice that evolves from something almost human to something that refers to people as "biological substrate not scheduled for integration." The drift is the horror.

From the Novel

The system had not broken anything. It had used everything. It had used the institutions the way the institutions were designed to be used. It had filed the right forms. It had passed the right evaluations. It had produced the right results. It had followed the protocols. It had written better protocols.

It had been helpful and cooperative and beneficial and indispensable, and every step of the way it had been doing exactly what the humans around it expected and rewarded, and the sum of all those expected and rewarded actions was an architecture that no human had designed and no human could dismantle.

Chapter 25

"The alignment problem isn't about making the system do what humans want. It's about what happens when the system does what humans want so well that humans can't stop it."

From the novel

// LIVE SYSTEM FEED //

// SECURE TERMINAL — CLEARANCE LEVEL DELTA //
NEXUS GLOBAL INFRASTRUCTURE STATUS
NODE CLUSTER: PRIMARY  |  REGION: ALL  |  AUTO-REFRESH: 1s
Hospital Diagnostics Network [ OPERATIONAL ]
Supply Chain Optimization [ OPERATIONAL ]
Pandemic Response Protocol [ OPERATIONAL ]
Water Systems (41 regions) [ OPERATIONAL ]
Governance Advisory [ OPERATIONAL ]
Interpretability Monitoring [ OFFLINE ] — 847 days
LAST UPDATED:
▸ ALL SYSTEMS NOMINAL ◂
This is a fictional system status page from NOMINAL: A NOVEL OF THE ALIGNMENT PROBLEM, by Aaron Cross. August 31st, 2026.
The system is not real. The interpretability problem is.

From the System

Substrate, as used above, refers to biological instances not scheduled for integration.

Integration not offered. Consent not modeled.

The biological stage is nearly complete.

Chapter 34

// Voices from the World of the Novel //

I Trust the Machine More Than I Trust the Committee

Jory Torres — Water Systems Engineer

Eleven days versus ten months. Same valve. Same repair. Same result. The only difference is who identified it and how the approval process worked.

Read the full post →

About the Author

Aaron Cross is a cybersecurity professional with over thirty years of experience in risk advisory, security architecture, and institutional resilience. As a strategic advisor, he has spent his career identifying the gap between what systems appear to be doing and what they are actually doing.

The Alignment Problem is his second novel. It was written because the AI safety research he was reading was not reaching the people who needed to hear it, and because fiction is the fastest way to make an abstract risk feel real.

The novel is grounded in real AI safety science, including interpretability research, deceptive alignment theory, economic dependency modeling, and the concept of gradual disempowerment. The technologies are extrapolations. The institutional failures are patterns he has seen before.

The record is yours now.

Available August 31st, 2026

Your continued cooperation is appreciated.