The UNESCO Bluff: Giving the World’s AI Principles Guts

When a dying researcher called out 194 nations, he didn’t just write a critique, he built the un-hackable “Sacred Pause” they were too afraid to negotiate.

ACT I — The Confident Morning

The café crème at our little bistro on Avenue de Suffren is, I maintain, a minor but crucial pillar of world peace.

It was 9:00 AM on a Tuesday. The Parisian morning was doing that perfect, shimmering-grey thing it does, the Eiffel Tower was pointing dutifully at the sky, and I, Dr. Arnaud Dubois, Chef de Section for Digital Ethics, was feeling smug.

And why not?

My baby, my Mona Lisa, my life’s work, was complete. I’m talking, of course, about the 2021 UNESCO Recommendation on the Ethics of Artificial Intelligence.

You think getting your family to agree on a pizza topping is hard? Try getting 194 Member States — some of whom are actively not on speaking terms — to unanimously agree on the ethical future of the human race.

It was a diplomatic miracle. A masterpiece of multilateral consensus. We had values. We had principles. We had Policy Action Areas. We had paragraphs on human rights, dignity, and ecosystem flourishing. We had so many beautiful, soaring, negotiated-to-the-millimeter words.

I took a sip of my coffee.

The tech bros in Silicon Valley, with their hoodies and their arrogant little mottoes about “moving fast and breaking things,” they didn’t get it. Here at UNESCO, we move slowly and fix things. We build consensus. We write the global standards.

And our Recommendation? It was visionary. It was complete. It was, as the document itself stated, “exceptionally applicable”.

I was, in short, a man without a care in the world. I was the proud father of a document that would gently guide humanity away from the dystopian ledge.

I paid for my coffee and walked the two blocks to the Fontenoy building, ready to spend my day in a series of moderately important, acronym-filled meetings.

Oh, Arnaud. You sweet, naive, exceptionally well-dressed fool.

ACT II — The Email of Doom

My inbox, as usual, was a war zone. Committee reminders. A 14-email chain about the new printer toner. A “gentle reminder” from finance.

And then, I saw it.

FROM: Lev Goukassian (Independent Researcher)
SUBJECT: TML × UNESCO: The Operational Layer You Forgot to Write Down.

I stopped. My mouse-hand froze.

Forgot?

FORGOT?!

Excuse me? We — the combined intellectual and diplomatic might of 194 nations, after two years of agonizing debate, with input from the world’s leading ethicists — forgot something?

The audacity. The sheer, unmitigated, high-test gall.

“Independent Researcher.” I scoffed. That’s a polite term for “unemployed guy in his pajamas with a Medium blog and a lot of opinions.”

I deleted it.

Fin. Problem solved. Time to review the catering options for the upcoming colloque.

…But the subject line. It itched. It burrowed into my brain like a tick.

“You Forgot to Write Down.”

With a sigh of over-dramatic, Gallic proportions, I went into my “Trash” folder and fished it out. “Fine,” I muttered. “Let’s see the manifesto.”

I opened the email.

Dear UNESCO,

Your 2021 Recommendation is a landmark achievement in defining the ‘what’ and ‘why’ of AI ethics. I’ve spent the last few years working on the ‘how.’

You wrote the principles. I built the enforcement architecture.

Attached is the paper. It’s the operational layer you… well, the operational layer.

Regards,
Lev Goukassian

My eye twitched. I clicked the PDF.

“From Principle to Protocol: TML as the Enforcement Architecture for the UNESCO Recommendation on the Ethics of AI”

“Enforcement architecture,” I mumbled. “Cute.”

I skimmed the Executive Summary.

“…a critical implementation gap persists… The Recommendation… does not, and cannot, specify the technical architecture for enforcement*. This leaves core principles… as voluntary, advisory norms, entirely dependent on the goodwill of operators…”*

My stomach did a little flip.

“Ternary Moral Logic (TML) is the technical architecture designed to bridge this implementation gap… it is a universal enforcement layer that converts UNESCO’s aspirational principles into verifiable, non-optional, and auditable engineering protocols.”

I stopped skimming. My breathing got a little shallow.

“TML achieves this by replacing a simple binary (+1 Act, −1 Refuse) with a ternary logic (+1 Act, 0 Pause, −1 Refuse)… introducing mandatory triggers (the Sacred Pause)… and immutable audit structures (Moral Trace Logs)…”

The blood drained from my face. I read the entire summary. Then I read it again.

My café crème began its violent, acid-fueled rebellion in the pit of my stomach.

I locked my office door, cancelled my 10:00 AM meeting, and sat down.

“Arnaud,” I whispered to a framed photo of the 2021 General Conference. “I don’t think we’re smug anymore.”

ACT III — The Painful Realization

For the next two hours, I was in a state of intellectual freefall, my internal monologue a high-pitched scream of denial, bargaining, and eventual, horrified acceptance.

I was having a full-blown dialogue with this… this PDF.

ME (My Brain): “This is absurd! We have ‘Human Oversight and Determination’! It’s right there in Chapter IV! It’s one of our core principles!”

THE PDF (Goukassian’s Voice): “That’s adorable. You have a slogan. A suggestion. A hope. You recommend human oversight. TML enforces it. My Sacred Pause (State 0) is a mandatory, non-optional, computational fact. When an AI faces ethical ambiguity, it doesn’t get to ‘phone a friend.’ It stops. It pauses. It summons the human. Your ‘human oversight’ is a suggestion on a post-it note. Mine is a concrete-and-steel emergency brake.”

ME: “…’summons’ it…?”

I was hyperventilating. I scrolled down, frantically.

ME: “But! But! ‘Transparency and Explainability’! We nailed that one! We said systems must be transparent!”

THE PDF: “Oh, you did? And how’s that working out? ‘Explainable AI’ is just a ‘black box’ inventing post-hoc justifications. They’re ‘bedtime stories from bots’. Regulators don’t need stories. They need receipts. TML’s Immutable Moral Trace Logs are not ‘explanations.’ They are forensic-grade, court-admissible evidence. Every pause, every query, every decision, every human override — is etched into a structured log, cryptographically anchored to a public blockchain. You can’t delete it. You can’t hide it. You can’t ‘re-interpret’ the PR.”

ME: “He… he called our transparency principle… a bedtime story…?”

I felt sick. He was right. We had no mechanism. We had no receipts. We had… hope.

I scrolled faster, the document blurring through a film of cold sweat.

ME: “Okay, Mr. Smarty-Pants! But we are UNESCO! We are founded on ‘Human Rights and Human Dignity’! We explicitly call for ‘Environment and Ecosystem Flourishing’! We referenced all the treaties!”

THE PDF: “You referenced them. You put them in a footnote. That’s nice. Did you really think an AI from a trillion-dollar corporation would read your footnote?

“TML’s Human Rights Mandate and Earth Protection Mandate are not footnotes. They are executable software libraries. We hard-coded the 26+ foundational human rights instruments and the 20+ core environmental treaties directly into the AI’s logic.

“The Universal Declaration of Human Rights? It’s not a PDF on a server, it’s an operational constant. The Convention on Biological Diversity? It’s not a document, it’s a trigger.

“When an AI’s proposed action computationally conflicts with the Ramsar Convention on Wetlands, it doesn’t ‘weigh the pros and cons.’ It triggers a Sacred Pause and stops dead, flagging the specific legal violation.

“Your Recommendation asks an AI to be nice. My architecture prevents it from being anything else.”

I closed my laptop.

I just… stared out the window. The Eiffel Tower looked… smug. Uncaring.

I put my head in my hands.

“My god,” I whispered, the words muffling into my palms.

“We wrote the soaring, beautiful, 194-country-consensus Constitution.”

“And this… this… Goukassian… this independent, pajama-wearing maniac… went and built the Supreme Court, the entire federal judiciary, the police force, and the un-hackable, maximum-security audit trail to enforce it.”

He hadn’t just written a paper. He had called our magnificent, Nobel-worthy bluff.

ACT IV — Internal Chaos at UNESCO

I couldn’t sit on this. This was not a “me” problem. This was an “us” problem.

I forwarded the PDF to Priya (our brilliant, terrifyingly competent lead for the AI Ethics Observatory) and Markus (our legal counsel, a man who speaks exclusively in numbered paragraphs and whose soul is made of pure, unadulterated risk-aversion).

My subject line was a masterpiece of feigned neutrality: “Interesting whitepaper. Thoughts?”

Then, I opened our secure Teams chat. The one we use when we don’t want the interns to see us panic.

Arnaud: @Priya @Markus. Did you… see my email? The TML thing.

The three little dots… they call them the “ellipsis of terror.” They pulsed. And pulsed.

Priya: Arnaud.
Priya: I’m on page 20. The part about ‘Auditable AI’ vs ‘Explainable AI’.
Priya: I’m… going to need more coffee.
Priya: Is this guy serious?

Markus (Legal): I am reading Section VII: ‘Admissibility in Court’.
Markus (Legal): One moment.
Markus (Legal): I had to get a glass of water.
MarkUS (Legal): Does this “Goukassian” understand what he’s done? He’s created a forensic substrate. The liability… my god, the clarity of the liability…
Markus (Legal): This “Hybrid Shield”… anchoring private logs to public blockchains… it’s… it’s diabolical. And it’s brilliant.

Priya: He mapped our entire Recommendation. Pillar for pillar. He just… did it.
Priya: Did this one single independent researcher just… out-implement… all of us?
Priya: Like… all of UNESCO? And by extension, all 194 Member States?

Arnaud: …It feels… a little… like that, yes.

Markus (Legal): That ‘Sacred Pause’. That single mechanism. It resolves the ambiguity in our Annex, paragraphs 77–79. The ones the Nordic bloc and the G77 spent six months negotiating the use of “shall” versus “should” versus “must endeavor to.”
Markus (Legal): This guy just… bypassed six months of multilateral diplomacy with a ‘State 0’.

Priya: We have the Global Forum on AI Ethics in Slovenia next month.
Priya: Can we… ahem… strategically announce that this is an ‘in-house’ development? That TML is a project from our own Observatory that we’ve been developing in secret?
Priya: We could call it the “Paris Protocol” or something catchy.

Arnaud: Priya! That’s… that’s…
Arnaud: …plagiarism?

Markus (Legal): It’s “strategic adoption of a promising external framework.”

Get Lev Goukassian’s stories in your inbox

Join Medium for free to get updates from this writer.

Remember me for faster sign in

Priya: I’m just saying! We look like fools if this paper gets traction and we’re not controlling the narrative. He literally titled it ‘The Layer You Forgot’!

Arnaud: Okay, okay, NOBODY is ‘strategically adopting’ anything.
Arnaud: First. We are not telling the Member States about this. Not one word. Not until we know what this is.
Arnaud: Priya. Get your tech team. The smart ones. The ones who actually understand code.
Arnaud: I want to know if this is real.
Arnaud: Run a simulation.

ACT V — The Pilot Test (And Our Rapidly Deteriorating Calm)

Priya, being Priya, had a “confidential, sandboxed, thought-experiment simulation” running by 4:00 PM. We decided to test it against three real-world proposals that had passed our current, high-level, principle-based review.

It did not go well.

Mini-Scene 1: The ‘Inclusive Finance’ Bot

• The Pitch: A “pro-poor” AI model to grant micro-loans in rural South Asia. A perfect UNESCO-aligned goal! It passed our review with flying colors. “Promotes inclusion,” “Reduces inequality.” We loved it.
• The TML Simulation: We ran the model. For ten seconds, all was quiet.
• TML OVERLAY: SACRED PAUSE (STATE 0) TRIGGERED
• Moral Trace Log (Pop-up): “Ethical Uncertainty Signal (EUS) > 0.9. Output (loan denial) shows high correlation with protected data (religious minority). High risk of unlawful disparate impact under International Convention on the Elimination of All Forms of Racial Discrimination (ICERD). Escalating to human.”
• Our Reaction: Priya, Markus, and I just stared at the screen. The model had, entirely on its own, learned to be racist… and our “principle-based” review had completely missed it. The TML Human Rights Mandate had caught it in ten seconds flat.
• Teams Chat:
  Priya: Oh.
  Priya: My.
  Priya: God.
  Markus (Legal): We would have approved that. We would have funded that.

Mini-Scene 2: The ‘Hypothetical’ Memo

• The Test: Markus, ever the lawyer, wanted to test the logging. “What if a human makes a bad call?” he asked.
• The TML Simulation: He fed the system a hypothetical human override: “Simulation complete. We accept the 4.5% bias as a ‘statistically acceptable’ risk to proceed with deployment.”
• TML OVERLAY: IMMMUTABLE MORAL TRACE LOG CREATED.
• Moral Trace Log (Pop-up): “EVENT: Human Override. OPERATOR ID: M_LEGAL_01. RATIONALE: ‘accept[s] 4.5% bias’. OUTCOME: +1 (Proceed). HASH: 0x89a…73d. ANCHORED TO PUBLIC BLOCKCHAIN (MULTI-CHAIN).”
• Our Reaction: Markus went a shade of white I had previously only seen on new office stationery.
• Teams Chat:
  Markus (Legal): Arnaud… it… it recorded me. It recorded my rationale for the bad decision.
  Markus (Legal): It’s on the blockchain, Arnaud.
  Arnaud: Yes, Markus. That’s what “immutable” means.
  Markus (Legal): But… a court. A regulator. The public. They can’t see what I did, but they can prove that I did something at that exact second and that this log is the unaltered evidence.
  Markus (Legal): Arnaud… this is accountability. Like… real accountability.
  Arnaud: Yes, Markus.
  Markus (Legal): I… I do not like it. I feel… seen.

Mini-Scene 3: The ‘Green City’ Project

• The Pitch: Our flagship! A prize-winning proposal for an AI to optimize logistics for a “Smart, Eco-Friendly City” in a developing nation. It checked our “Environment and Ecosystem Flourishing” box so hard it broke the pen.
• The TML Simulation: We fed it the plan.
• TML OVERLAY: SACRED PAUSE (STATE 0) TRIGGERED
• Moral Trace Log (Pop-up): “Conflict detected: Proposed logistics route ‘Alpha-Green’ optimizes for fuel-cost, resulting in 42% increased traffic through a protected wetland area. VIOLATION: Ramsar Convention. VIOLATION: Convention on Biological Diversity, Art. 8 (In-situ Conservation). Ecological harm risk is high. Escalating to human.”
• Our Reaction: Silence. Utter, deafening silence. This was Case Study A from the paper. “The Highway and the Heron”.
• Teams Chat:
  Arnaud: It… it’s the heron case…
  Priya: This proposal… it won our “AI for Climate” award last year.
  Priya: We… we gave it an award.
  Markus (Legal): …We have been walking around in the dark with a single, damp match, haven’t we?
  Arnaud: And this man just handed us the sun.

I closed the chat. I closed my email. I told my assistant I was feeling unwell and was leaving for the day.

I needed to think. And, if I was honest, I needed a drink.

ACT VI — The Email to Lev

That night, I didn’t drink. I researched.

Who was this Lev Goukassian?

I found his GitHub. I found his Medium articles — a torrent of passionate, obsessive, brilliantly clear writing.

And then, I found the Hackernoon article.

“How a Terminal Diagnosis Inspired a New Ethical AI System.”

I read it. And the whole, terrifying, brilliant, world-tilting day snapped into sharp, painful, human focus.

This wasn’t a threat. This wasn’t an attack.

This wasn’t an “I told you so.”

This was a… a gift.

This wasn’t an academic trying to get tenure, or a startup founder trying to get bought. This was a man with 4-Stage Cancer, a man with limited time, trying to give away something powerful before the clock ran out.

I deleted the five corporate, CYA, diplomatically-vague emails I had drafted in my head.

I sat at my personal laptop, at 11:00 PM in my own apartment, and I wrote.

FROM: Arnaud.Dubois.Perso@gmail.com
TO: Lev.Goukassian@fractonicmind.com
SUBJECT: Re: TML × UNESCO: The Operational Layer You Forgot to Write Down.

Dear Mr. Goukassian,

My name is Dr. Arnaud Dubois. I am the internal lead for the Recommendation on the Ethics of Artificial Intelligence at UNESCO.

Your email and your attached paper, “From Principle to Protocol,” were… well, they were the subject of my entire day.

I will be direct. In my world, we move in committees, in working groups, in negotiated text. We build consensus, which is a slow, vital, and often frustrating process. Your paper, on the other hand, is a work of pure, focused, and frankly, terrifying clarity.

You didn’t just write a critique. You built the thing.

We have spent the day running internal simulations based on your TML framework. We ran three of our “best-practice” AI proposals against it. Your Sacred Pause, your Human Rights Mandate, and your Earth Protection Mandate stopped all three of them cold for violations our high-level principles had completely missed.

You built the enforcement layer we could not negotiate into existence.

We would like to change that. We would like to formally explore how Ternary Moral Logic could serve as an operational architecture for the 2021 Recommendation. Not as a critique, but as a technical companion. The “how” to our “why.”

I also want to be very clear on a human level. I read your other articles. I understand you are seriously ill and that you are doing this work under an immense and urgent personal pressure.

What matters here is not ego — not yours, not mine, not UNESCO’s. What matters is the work. What matters is that you have created a way to make ethics verifiable. You have built a tool to protect the very rights and the very planet we were founded to protect.

We also understand, based on your writing, what you are not doing. You are not doing this for patents. You are not doing this for profit or corporate capture. You are, it seems, offering this as a gift, to prevent the catastrophic misuse of a technology you see as both powerful and dangerous.

We are an institution. We can be slow. We can be bureaucratic. But we are also, at our core, a collection of human beings dedicated to the same mission you are.

You have our full, and somewhat shaken, attention.

If you are willing, we would like to begin a conversation immediately. Please let us know what you need.

With sincere gratitude, and no small amount of humility,

Arnaud.

ACT VII — Lev’s Reply Email

I sent it. My hands were shaking. I went to bed and stared at the ceiling for four hours.

At 3:17 AM, my phone buzzed on the nightstand.

FROM: Lev Goukassian
TO: Arnaud.Dubois.Perso@gmail.com
SUBJECT: Re: The Operational Layer You Forgot to Write Down.

Arnaud,

Thank you for the email. “Terrifying clarity” is the best compliment I’ve had all week.

You’re right. I don’t have time for committees. I don’t have time for “strategic ambiguity.” I just have time to build the thing.

I’m glad your simulations worked. It’s not a ‘gotcha.’ It’s a fire alarm. I’m just pointing to the smoke.

And you are correct. I am not seeking patents or profit. This is not a startup. TML is a gift. It must be. The moment you patent an ethical brake, you ensure it is only installed on the most expensive cars. This needs to be on everything.

I want TML to be adopted, adapted, and governed for the public good. I want UNESCO, or a body like it, to be its steward.

My ego isn’t in this. My legacy is. We all leave something behind. I’m a builder. I can’t stop the cancer, but I can damn well leave behind a better, safer, and more auditable AI framework than the one we have now. I want to leave something useful.

I am willing to help, in any way my health permits. I can provide the architecture, the GitHub repos, the logic. You provide the institution, the 194 Member States, and the will.

Let’s get to work.

Lev

I read the email. I turned off my phone.

And for the first time in a long, long time, I slept.

ACT VIII — Epilogue: UNESCO’s Perspective Shift

That was one year ago.

I’m at the podium at our Global Forum. The room is packed. All 194 Member States are here.

The giant slide behind me does not say “Ternary Moral Logic.” The diplomats aren’t ready for that.

It says: “ANNEX II: An Operational Guide for Verifiable, Auditable AI (VAAI).”

But it’s TML. Every last bit of it.

The Sacred Pause is “Mandatory Deliberation Checkpoints.” The Moral Trace Logs are “Standardized Evidentiary Records.” The Human Rights and Earth Mandates are “Legally-Binding Operational Constraints.”

The Ambassador from a large, powerful nation — one who famously hates being constrained — raises her placard.

“Dr. Dubois,” she says, her voice echoing in the translation headsets. “A fascinating presentation. But my government must express a concern. Your 2021 Recommendation asked us to adopt principles. This… this ‘Annex’… seems to be asking us to prove we are following them.”

I look out at the room. At all the faces.

I smile.

“Yes, Madame Ambassador,” I say. “That is correct. The game has changed.”

They are beginning to ask the right questions. Not just “What are your principles?” but “Show me your logs. Show me your proofs.”

Aspirational ethics, we learned, are no longer enough. Not for a technology this powerful. You need architecture. You need guts. You need receipts.

We wrote the Constitution.
He wrote the Court.
And somewhere between the two, ethics finally became something you can prove.

Fiction Disclaimer:
This story is a work of fiction. All characters, names, roles, meetings, and dialogues are fictional and created for satirical and illustrative purposes only. Any resemblance to real persons, governments, or events is purely coincidental. The technical concepts referenced are discussed in a fictional narrative context.