The Ecocide Precedent: Why Tech, AI and Financial Executives Must Face Personal Criminal Liability

Last Updated: September 8, 2026By

Europe Decided That Destroying a Forest Can Send a Man to Prison. Now Do It for the Men Destroying Everything Else.

A case for personal criminal liability for technology, artificial intelligence, and financial executives — in every jurisdiction that still calls itself a democracy.

By Vahid Razavi NoEthicsInBigTech.com

I. What Europe Actually Did

On 11 April 2024, the European Parliament and Council adopted Directive (EU) 2024/1203 on the protection of the environment through criminal law. It entered into force on 20 May 2024. Member states were required to write it into national law by 21 May 2026.

Read what it says, because the words matter more than the summary.

It creates a category the drafters called a “qualified offence” — conduct that European lawyers describe openly as comparable to ecocide. Article 3(3) defines it as an offence that causes the destruction of, or widespread and substantial damage which is either irreversible or long-lasting to, an ecosystem of considerable size or environmental value, or to a habitat within a protected site, or to the quality of air, soil or water.

And then it does the thing that no digital regulation in the world has been willing to do.

It sends people to prison.

  • A maximum penalty of at least ten years’ imprisonment for offences that cause the death of any person.
  • A maximum of at least eight years for the qualified “ecocide” offences.
  • For companies: fines of at least 5 percent of total worldwide turnover, or €40 million, whichever is greater, for the most serious offences. At least 3 percent or €24 million for the rest.
  • A limitation period of at least ten years, running from the date the offence was committed or from the date the authorities discovered it.

Note who is exposed. The law firm Crowell & Moring, writing to its corporate clients, put it plainly: because criminal conduct under the directive includes omissions and failures to act, a failure by a CEO or a board member — a failure to stop, to monitor, to prevent — may incur criminal liability in their personal capacity.

Belgium went further and went first. On 22 February 2024, five days before the European Parliament’s final vote, Belgium’s Federal Parliament adopted a new Criminal Code containing, at Article 94, a domestic crime of ecocide: deliberately committing, by act or omission, an illegal act causing serious, widespread and long-term damage to the environment, in the knowledge that this act is causing such damage. Belgium became the first EU member state to criminalise ecocide, and the first to recognise it at the international level alongside genocide, war crimes, crimes against humanity, and the crime of aggression. It applies to individuals in the highest positions of decision-making power, and to corporations, in peace and in war alike.

And on 9 September 2024, three Pacific island states — Vanuatu, Fiji, and Samoa — formally submitted a proposed amendment to the Rome Statute of the International Criminal Court to make ecocide the fifth international crime. Their definition, drawn from the 2021 Independent Expert Panel convened by the Stop Ecocide Foundation and chaired by Philippe Sands KC and Dior Fall Sow, reads:

unlawful or wanton acts committed with knowledge that there is a substantial likelihood of severe and either widespread or long-term damage to the environment being caused by those acts.

Sit with that standard for a moment. Knowledge that there is a substantial likelihood. Not certainty. Not intent to destroy. Not a signed confession. Knowledge, plus substantial likelihood, plus severity — and the decision to proceed anyway.

Hold that sentence. We are going to need it again in about nine hundred words.

II. The Sentence That Should End the Debate

I want to be exact about this, because I have spent years watching arguments get won by a single verified quotation and lost by a single exaggerated one.

In 2015, speaking with then-Airbnb CTO Mike Curtis at Airbnb’s Open Air conference — months before he co-founded OpenAI — Sam Altman said:

“AI will probably most likely lead to the end of the world, but in the meantime, there’ll be great companies.”

That is the quote. It is real, it is on the record, and it has been reported and re-reported for eleven years. His defenders say it was a joke. Perhaps it was. I have listened to men in Silicon Valley make that particular kind of joke for three decades, and I have learned something about them: the joke is where they put the thing they actually believe, so they can say it out loud without being held to it.

But we do not have to rely on a 2015 aside, because the numbers came later, and they came from the men in charge.

On 30 May 2023, the Center for AI Safety published a one-sentence statement:

“Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.”

It was signed by Sam Altman of OpenAI. By Dario Amodei of Anthropic. By Demis Hassabis of Google DeepMind. By Ilya Sutskever. By Geoffrey Hinton and Yoshua Bengio, the two most-cited computer scientists alive. Hundreds of signatories in total.

Then came the percentages.

On 17 September 2025, at the Axios AI+ DC Summit, Axios co-founder Jim VandeHei asked Dario Amodei — the chief executive of Anthropic — what probability he would assign to AI ending in disaster. Amodei answered: “I think there’s a 25% chance that things go really, really badly.” He added that he prefers to focus on the 75 percent chance that it goes well.

Elon Musk has put his own number at as high as 20 percent. Sundar Pichai, chief executive of Google, told Lex Fridman that he is optimistic about the doom scenarios, but that “the underlying risk is actually pretty high.”

So let me state the position of this industry as its own leaders have stated it, with no editorial embellishment:

The men building this technology estimate somewhere between a one-in-five and a one-in-four chance that it ends in catastrophe for the human species. They have said so publicly, by name, with their titles attached. And they are shipping anyway.

Now go back and read the Vanuatu definition again. Unlawful or wanton acts committed with knowledge that there is a substantial likelihood of severe and either widespread or long-term damage.

Twenty-five percent is a substantial likelihood. It is not a fringe estimate from a critic. It is the CEO’s own number, given voluntarily, on a stage, to a journalist, in Washington, D.C.

If a chemical executive stood on a stage in Washington and said there was a 25 percent chance his product would poison a continent, and then shipped it the following Tuesday, the Environmental Crime Directive would already have his name on a docket.

III. The Bunker Is Evidence

I have written before that these men must not be free to retreat to bunkers in Hawaii or Patagonia while their tools fuel slaughter. I want to sharpen that now into something a prosecutor can use.

The bunker is not a lifestyle choice. The bunker is a document. It is a contemporaneous written record of foreseeability, executed in concrete, at enormous cost, by the defendant himself.

Consider the record.

Mark Zuckerberg has been building Koolau Ranch on Kauai since 2014 — roughly 1,400 acres, with an underground shelter of about 5,000 square feet with its own energy and food supplies, connected to the main estate by a tunnel. Workers signed non-disclosure agreements. A six-foot wall shielded the site. Permits referred to “basements.” Total projected cost has been reported north of $270 million. He has separately spent around $110 million on eleven properties in Crescent Park, Palo Alto, adding roughly 7,000 square feet of underground space. Asked directly whether it is a doomsday bunker, he called it “just a little shelter,” like a basement.

Peter Thiel obtained New Zealand citizenship in 2011 after spending twelve days in the country, bought a 477-acre estate at Wanaka, and applied to build a hillside bunker compound at Queenstown. The council rejected the application in 2022.

Sam Altman told The New Yorker’s Evan Osnos, for the January 2017 piece “Doomsday Prep for the Super-Rich,” that his contingency plan involved flying with Thiel to that New Zealand property, and described a go-bag with guns, gold, antibiotics, potassium iodide, batteries, water, and gas masks.

Reid Hoffman of LinkedIn coined the phrase for it — “apocalypse insurance” — and estimated that more than half of Silicon Valley billionaires have some.

And Ilya Sutskever, OpenAI’s former chief scientist, reportedly told colleagues, in a line recorded in Karen Hao’s book Empire of AI: “We’re definitely going to build a bunker before we release AGI.”

Before we release it.

I am not interested in mocking a rich man’s basement. I am interested in what it proves. In every criminal system on earth, the hardest element to establish is the defendant’s state of mind. What did he know? When did he know it? Prosecutors spend years and millions reconstructing it from emails and testimony and inference.

These men have saved us the trouble. They have built the mens rea out of steel and poured it into a hillside on Kauai. You do not buy a blast door for a risk you consider negligible. You do not put your builders under NDA for a basement. You do not say before we release it unless you have already decided that releasing it is the thing that makes the shelter necessary.

The bunker is the confession. The only thing missing is a jurisdiction willing to read it.

IV. The Gap, in One Comparison

Here is the contrast that should be printed and taped to the door of every ministry of justice in Europe.

Environmental harm. Directive (EU) 2024/1203. Ten years’ imprisonment for a natural person. Eight years for ecocide-comparable conduct. 5 percent of worldwide turnover or €40 million for the company. Transposed into national criminal law by May 2026. Personal liability for CEOs and board members, including for omissions.

Existential technological harm. Regulation (EU) 2024/1689 — the AI Act. Maximum penalty: €35 million or 7 percent of global turnover for prohibited practices. €15 million or 3 percent for high-risk violations. No imprisonment. No natural persons. No named defendants. Nobody’s door gets knocked on.

And then it got weaker. On 24 July 2026 the EU published Regulation (EU) 2026/1744, the Digital Omnibus on AI. It entered into force on 27 July 2026 — six days before the AI Act’s original high-risk compliance deadline of 2 August 2026. The Annex III high-risk obligations were deferred to 2 December 2027. The product-embedded high-risk obligations to 2 August 2028.

Read that sequence again. The most consequential obligations in the world’s first comprehensive AI law were pushed back by roughly sixteen months, six days before they were due to bite.

America is no better. California’s SB 53, the Transparency in Frontier Artificial Intelligence Act, signed by Governor Newsom on 29 September 2025 and effective 1 January 2026, was genuinely first in the nation — and it caps civil penalties at one million dollars per violation. New York’s RAISE Act, signed 19 December 2025 and finalised by chapter amendment on 27 March 2026, effective 1 January 2027, was negotiated down to match California’s revenue thresholds. Eight days before Hochul signed, President Trump issued an executive order directing the Department of Justice to challenge state AI laws inconsistent with a “minimally burdensome” national standard.

One million dollars. To a company whose CEO says there is a one-in-four chance of civilisational catastrophe.

I want to be very clear about what a fine is to these companies. I have made this argument about the Digital Markets Act and I will make it again here, because nothing has changed. A regulatory fine is a business expense. It is forecast, provisioned, appealed for years, and ultimately paid from a fund that is a rounding error on an annual earnings statement. Amazon, Google, and Microsoft have collectively paid billions in European fines and their market power has not moved by a percentage point. The fines are not deterrence. They are a licensing fee for continued impunity.

A criminal prosecution is a different thing entirely. It attaches to a person. It cannot be paid by shareholders. It cannot be indemnified into oblivion. It follows the defendant through every border crossing, every conference invitation, every bank compliance review, for the rest of his life.

That is why they will fight this. And that is exactly why it works.

V. This Is Not a New Idea. It Is an Old One We Stopped Using.

Every objection to what I am proposing has been made before, in a courtroom, by a defendant who lost.

Hamburg, March 1946

Bruno Tesch owned Tesch & Stabenow, a firm that distributed Zyklon B — a pesticide. Karl Weinbacher was his authorised deputy. Neither had ever been a member of the German government or armed forces. Neither operated a gas chamber. Neither killed anyone with his own hands.

The charge before the British Military Court was that they had knowingly supplied a commodity to a branch of the state that was using it for mass extermination, and had continued to do so, in increasing quantities, after they knew. By early 1944 the consignment to Auschwitz was approaching two tons a month. They had also, the evidence showed, supplied it stripped of the warning odorant normally added to protect humans from accidental exposure.

The defence argued that they only supplied the means. That commerce is not killing. That the customer’s use is the customer’s business.

The court convicted Tesch and Weinbacher on 8 March 1946. They were executed at Hameln on 16 May 1946. A third defendant, Joachim Drosihn, the firm’s gassing technician, was acquitted — because he had no insight into corporate policy and no power to change it.

That last detail is the whole architecture of the doctrine I am asking legislatures to revive. The engineer who cannot see the policy is not the defendant. The executive who sets it is.

I make this comparison with full awareness of its weight, and I do not make it loosely. I am not saying an AI executive is a Nazi. I am saying that the legal principle established at Hamburg — that knowingly supplying an instrument to those you know are using it for mass harm is itself a crime, committed by the supplier, in his own person — is a principle of international law that has never been repealed. It has only been allowed to fall out of use against the powerful.

Paris, 13 April 2026

Eighty years later, the same principle came back.

The Paris Criminal Court convicted the French cement company Lafarge SA and eight individuals — including four former top executives — of financing terrorist organisations and violating international sanctions, over payments made to armed groups including the Islamic State to keep a Syrian plant running in 2013 and 2014.

Former chief executive Bruno Lafont was sentenced to six years in prison, effective immediately. Christian Herrault, the former deputy director for Syria operations, received five years and a €225,000 fine, with immediate imprisonment ordered. Sentences across the eight ranged from eighteen months to seven years. The company received the maximum available fine of €1.125 million on the terrorism count, a joint customs fine of €4.57 million, and confiscation of €30 million in assets.

Two things about that judgment matter enormously for what I am arguing.

First, the court pierced the corporate veil. Lafarge argued that the parent could not be liable for its Syrian subsidiary, separate legal entities, standard group structure. The court rejected it, finding that Lafarge exercised effective control operationally and functionally, not merely as a shareholder, and describing the subsidiary — in the words of one of Lafarge’s own executives — as a “paper structure.”

Second, the doctrine on knowledge is already settled. On 7 September 2021, in the same case, the Cour de cassation held that one can be complicit in crimes against humanity even without the intention of being associated with the crimes committed. Knowingly paying millions to an organisation whose purpose is criminal suffices — regardless of whether the party concerned was acting to pursue a commercial activity.

There is the answer to “we were only selling a commercial service.” It is not a novel argument I am inventing. It is the binding holding of the highest court in France, applied and affirmed, and a former CEO went to prison under it four months ago.

Reykjavík, 2010–2018

And for anyone who says that financial executives simply cannot be prosecuted for systemic harm: Iceland did it. Following the 2008 collapse, Icelandic courts sentenced dozens of bankers to prison. By 2018, local reporting tallied 36 bankers sentenced to a combined 96 years. Kaupthing’s former CEO Hreiðar Már Sigurðsson received the heaviest sentence. All of the bank directors went to jail.

Iceland has a population smaller than Fresno. It did what the United States, with the largest justice apparatus on earth, declined to do even once at the executive level.

It was never a question of capacity. It was always a question of will.

VI. The Bankers Are Not Bystanders

Everything I have said about technology executives applies with equal force to the people who fund them, and I want to be specific about why, because the financial sector has spent fifteen years perfecting the argument that capital is morally inert.

It is not. Capital is a decision. Every allocation is a choice made by a named human being with a title and a compensation package, on the basis of a risk memorandum that exists in writing.

Three categories of conduct should carry personal criminal exposure:

One — financing with knowledge. This is Lafarge, exactly. If an investment committee approves a facility, a round, or a compute contract for a venture whose principals have publicly estimated a substantial likelihood of catastrophic harm, and that committee proceeds, the Cour de cassation’s holding applies on its own terms: commercial motive goes to motive, not to intent. Knowledge plus material assistance is enough.

Two — privacy and surveillance violations at scale. When a financial institution acquires, brokers, or monetises population-scale personal data in violation of data protection law, the current sanction is a fine assessed against a legal person. That is not accountability, it is arithmetic. The officer who signed the data-sharing agreement should be exposed in the same way that the officer who signs off on an illegal discharge into a river is now exposed under Directive 2024/1203.

Three — the risk that was written down and ignored. Every institution of any size maintains a risk register. Every one of them documents, in writing, the harms it has identified and declined to mitigate. That document is either a compliance artefact or it is evidence. Under the regime I am proposing, it becomes evidence — and its destruction becomes a separate offence, exactly as the UK Online Safety Act already makes it an offence for a senior manager to suppress, destroy, or alter information sought by the regulator.

The corporate shield was invented to encourage productive risk-taking by limiting financial exposure. It was never intended, in any jurisdiction, to insulate a natural person from criminal responsibility for a decision he personally made. Somewhere in the last forty years, the two got confused, and the confusion has been extremely profitable for a very small number of people.

VII. What the Law Should Say

I am not asking for a resolution. I am not asking for a framework, a set of principles, a voluntary commitment, or another summit. Those exist in abundance and they have produced a $1 million cap and a sixteen-month deferral.

I am asking legislatures to copy a statute that already exists, change the nouns, and pass it.

Here is the model.

Offence 1 — Reckless deployment of a high-capability system

Adapting the Vanuatu/Fiji/Samoa language directly:

Unlawful or wanton acts of development, release, or deployment of an artificial intelligence system or comparable technology, committed with knowledge that there is a substantial likelihood of severe and either widespread or long-term harm to human life, health, critical infrastructure, or democratic institutions.

Penalty: a maximum of at least ten years’ imprisonment for natural persons. Matching the death-causing tier of the Environmental Crime Directive, because the harm contemplated is not comparable to a chemical spill — it is, by the defendants’ own estimates, comparable to a pandemic or a nuclear exchange.

Offence 2 — Failure of a responsible officer to prevent

The United Kingdom has already built this and it took effect two months ago. Section 250 of the Crime and Policing Act 2026, in force from 29 June 2026, extends the “senior manager” attribution model — previously limited to economic crime under section 196 of the Economic Crime and Corporate Transparency Act 2023 — to all criminal offences. Where a senior manager commits an offence within the actual or apparent scope of their authority, the organisation is itself guilty of that offence. The definition is functional and role-based: divisional heads, regional leaders, senior executives, whether or not they sit on the board.

The Online Safety Act 2023 supplies the reverse direction: where the company commits certain offences and a senior manager has failed to take all reasonable steps to prevent it, that senior manager commits an offence personally.

Both directions should apply to frontier technology. Neither requires inventing anything.

Offence 3 — Financing or investment with knowledge

Codify Lafarge. Knowingly providing capital, compute, or infrastructure to an enterprise where the provider knows of a substantial likelihood of severe harm constitutes complicity, and commercial purpose is not a defence.

Corporate penalties

At least 5 percent of total worldwide turnover, or €40 million, whichever is greater. Not “up to.” The Environmental Crime Directive sets a floor, and the floor is what makes it real.

The consequences that actually bite

  • Director disqualification across all participating jurisdictions, mutually recognised.
  • Clawback of compensation, equity, and bonuses earned during the offence period.
  • Prohibition on indemnification. No company may indemnify an officer, or purchase insurance covering an officer, against criminal liability under these provisions. If this is not written explicitly into the statute, D&O policies will absorb it within one renewal cycle and the deterrent will evaporate.
  • Personal asset freezing on charge, not on conviction.

Evidence preservation — the provision that decides whether any of this works

Mandatory retention, for a minimum of ten years matching the directive’s limitation period, of: internal risk assessments, red-team and evaluation results, safety-case documentation, decisions to override or defer a safety recommendation, and any internal estimate of catastrophic risk. Destruction, alteration, or suppression is a separate offence carrying its own custodial sentence, on the Online Safety Act model.

I put this near the end of the list and I want to move it to the top of your mind. Every accountability regime that has failed in my lifetime failed here. The documents are the case. Preservation is the first order a prosecutor should seek, before anything else, in any jurisdiction, on day one.

Jurisdiction — the part that removes the bunker

This is the provision that answers the question I am asked more than any other: what stops them from simply leaving?

  • Territorial jurisdiction where any element of development, deployment, or effect occurs in the territory.
  • Effects jurisdiction where harm is suffered by residents, regardless of where the decision was made.
  • Habitual residence and presence jurisdiction over any officer resident in or present within the territory.
  • Subsidiary directors. This is the practical key and it is why the target set must be reframed. The directors and legal representatives of European subsidiaries live inside the European Arrest Warrant zone. A warrant against a Dublin-, Amsterdam-, or Paris-resident director is immediately enforceable in the issuing state. You do not need extradition from Seattle to change the behaviour of a board.
  • Mutual recognition of disqualification orders and asset freezes, so that a finding in one jurisdiction is not a local inconvenience but a global condition.

Let me be honest about the limits, because this document is meant to be usable by lawyers and I will not overstate its reach. A European Arrest Warrant has no effect in the United States, and the United States does not extradite its nationals on charges of this character. Interpol Red Notices against American technology executives over politically charged conduct would attract an immediate Article 3 challenge with a substantial chance of success, and building a strategy around them signals inexperience.

But an EAW makes the named person arrestable the moment he sets foot in any of the twenty-seven member states, and arrestable at any external Schengen border, because it circulates through the Schengen Information System. For senior executives of global companies with major European operations, European regulatory exposure, and European customers, that is a genuine and permanent constraint.

It means no Davos. No keynote in Paris. No honorary degree in Berlin. No yacht in the Mediterranean. No transit through Frankfurt. And no bunker in Queenstown, because New Zealand is a participant in the international legal order and the Queenstown Lakes District Council has already demonstrated that it is not for sale.

That is the point. Not a perp walk in Seattle. A world that gets smaller every year, for men who have spent their lives assuming it would only get larger.

And at the international level

Support the Vanuatu, Fiji, and Samoa amendment to the Rome Statute. Not only because ecocide belongs there — it does — but because its adoption establishes the precedent that the list of international crimes is not closed, that Article 5 can be amended, and that a category of harm defined by knowledge of substantial likelihood is justiciable before an international court. Article 5 has been amended once before, for the crime of aggression at Kampala. It can be amended again.

Note the structural limit honestly: Article 25(1) of the Rome Statute confines the ICC’s jurisdiction to natural persons. It cannot reach the companies.

For this campaign, that is not a weakness. It is the entire objective.

VIII. Where This Gets Filed First

The argument is worthless without venues. In descending order of realism:

Belgium. Already has ecocide in its criminal code, already has general corporate criminal liability, already the first mover in Europe, and already comfortable with the language of international crimes. The natural legislative home for the first “reckless deployment” statute.

France. Lafarge is French law. The Cour de cassation’s holding on complicity is binding domestic precedent, the Paris Criminal Court has now imposed immediate custodial sentences on a former CEO under it, and Article 113-2 of the Code pénal keeps the plainte avec constitution de partie civile route open when a prosecutor declines.

The Netherlands. Corporate criminal liability is available, infrastructure and data are physically present on Dutch soil, and Article 12 of the Wetboek van Strafvordering provides a judicial remedy when a prosecutor refuses to act. Evidence preservation orders should be sought here first.

Ireland. The EMEA contracting entities for the largest platforms are Irish companies with Irish-resident directors. Section 8 of the International Criminal Court Act 2006 is the best-fitting statutory text in Europe. The Director of Public Prosecutions gate is real and should be planned for, not wished away.

Spain. Continuing, via the export-control route under Real Decreto-ley 10/2025 of 23 September 2025, and on the basis that dual-use controls under Regulation (EU) 2021/821 already extend to software, technology, intangible transfers, and technical assistance — including cyber-surveillance items. UN Special Rapporteur Francesca Albanese expressly recommended in A/HRC/59/23 (30 June 2025) that militarised AI, cloud, and surveillance technologies be classified as dual-use.

Five jurisdictions. Five filings. None of them requires a single new treaty.

IX. The Objections

“This will chill innovation.”

The Environmental Crime Directive did not end European manufacturing. Sarbanes-Oxley did not end American accounting. What personal liability chills is not innovation — it is the specific decision to ship a thing you privately believe might kill people, because the bonus vests in March. If your business model cannot survive the possibility that you personally answer for the harm it causes, you do not have a business model. You have an externality with a marketing department.

“You cannot prove causation for a hypothetical future harm.”

You do not have to. Read the standard again: knowledge that there is a substantial likelihood. This is a conduct offence, not a results offence. The environmental version does not require you to wait for the ecosystem to die. Neither does this one. And the reckless endangerment model — punishing the creation of unreasonable risk rather than the materialisation of it — has existed in every common law and civil law system for more than a century. There is nothing exotic here.

“The corporate veil protects legitimate business.”

It protects against financial exposure. It has never, in any jurisdiction, been a shield against criminal responsibility for one’s own acts. Tesch owned a company. Lafont ran one. The Paris court looked at Lafarge’s group structure in April and called the subsidiary a paper structure, because that is what one of Lafarge’s own executives called it.

“This is extraterritorial overreach.”

The AI Act already applies to any provider whose systems touch anyone in the European Union, regardless of where the company is headquartered. Nobody called that overreach. They called it Tuesday. All I am asking is that the same reach be extended from the fine to the person.

“We are just a platform. We just sell infrastructure.”

Bruno Tesch sold pesticide. Lafarge sold cement. In 1946 and again in 2021 and again in April 2026, courts held that knowingly supplying the means, to those you know will use them for grave harm, is complicity in that harm — and that a commercial purpose does not convert it into innocence.

That argument has now lost three times in eighty years. It is not a defence. It is a habit.

X. Why This Is Not Less Than a Crime Against Humanity

I want to close where I began this work, because everything in this article rests on a single distinction.

There are two ideas competing for the future, and everything else is commentary.

The first is I am one. I am one, my shareholders are one, my quarterly earnings are one, my compound on Kauai is one. Under this idea, a 25 percent chance of civilisational catastrophe is a risk to be personally hedged — with a tunnel, a food supply, a second passport, a go-bag with gold and gas masks. The hedge is rational. It is even, in its own terms, responsible. It simply does not include you.

The second is We are one. Under this idea, a 25 percent chance of catastrophe is not a risk to be hedged. It is a decision that no individual has the standing to make on behalf of eight billion people who were never asked.

That is the crime. Not the technology. The unilateral assumption of the authority to gamble with everyone.

Crimes against humanity have never been defined by the weapon. They are defined by scale, by knowledge, and by the deliberate treatment of human beings as inputs to someone else’s objective. A man who releases a system he believes has a one-in-four chance of ending the species, having first built himself a shelter, has done every one of those things. He has assessed the scale. He has documented the knowledge. And he has treated the rest of us as an acceptable variance in someone else’s return profile.

The environmental movement spent fifty years making this argument about rivers and forests, and in 2024 it won. Ten years. Five percent of worldwide turnover. Forty million euros. Personal liability for CEOs and board members, including for what they failed to do.

We are not asking for anything Europe has not already granted to a wetland.

XI. What I Am Asking For

To legislators — in Brussels, Madrid, Paris, The Hague, Dublin, Brasília, Ottawa, Canberra, Seoul, Nairobi, Wellington, and every capital that has ever complained about being a rule-taker in the digital economy: take Directive (EU) 2024/1203, change “ecosystem” to “human life, health, critical infrastructure, and democratic institutions,” and pass it. The drafting is done. It survived four years of trilogue negotiation. It has a transposition record. Copy it.

To prosecutors — the doctrine you need already exists in your own jurisprudence. In France it is Lafarge. In international law it is Hamburg 1946. In Europe it is Article 3(3). You are not being asked to invent a theory. You are being asked to apply an existing one to defendants who are wealthier than the ones you usually apply it to.

To the engineers, researchers, and safety staff inside these companies — you are Drosihn, and Drosihn was acquitted. The court found he had no insight into corporate policy and no power to alter it. That acquittal is your protection and it is also your assignment. What you know, and what you can see from where you sit, is the case. Preserve it. And when the moment comes, as Daniel Ellsberg said, tell the truth, take the risk, and say what you know when you know it.

To everyone else — sign the petition at NoEthicsInBigTech.com. Read the documented record. Share this with a legislator, a journalist, a prosecutor, a law student, or a pension fund trustee who has never once been asked whether the companies in the portfolio have named a probability of human extinction on a public stage.

Evil men and their machines do not prevail. History is long, and it is littered with their kind.

But history does not bend toward justice by itself. It bends because people push it. Because petitions get signed and cases get filed and courts get petitioned, and because on 13 April 2026 a former chief executive walked into a courtroom in Paris expecting a fine and walked out into custody.

That happened. It happened this year. It happened because a handful of former Syrian employees and two NGOs filed a complaint in 2016 and refused to go away for ten years.

Ten years is not long. The men in the bunkers are counting on it being longer than our attention span.

Let us disappoint them.

Live in peace, not just rest in peace.

ForeverPeaceNow.

SOURCES

Every source below was retrieved and verified. Where a claim could not be verified to a primary or authoritative secondary source, it has been removed from the text rather than softened.

The European ecocide framework

Belgium

The ICC ecocide amendment

Statements by AI executives

The bunkers

The AI Act and its deferral

US frontier AI law

The Zyklon B case

The Lafarge case

Iceland

UK senior manager liability

  • No Ethics in Big Tech (2026), Vahid Razavi — NoEthicsInBigTech.com/Book/
  • Ethics in Tech and Lack Thereof (2018), Vahid Razavi
  • Forever Peace Now (documentary, 2025) — ForeverPeaceNow.com
  • MyAWSStory.com | ParentsPlea.com | ConflictTour.com | BigTechSellsWars.com
  • Video archive: youtube.com/@EthicsInTechnology | youtube.com/@bizcloud | archive.org/details/@vahidr

Vahid Razavi is the author of “The Age of Nepotism”, “No Ethics in Big Tech” and “Ethics in Tech and Lack Thereof”, and the director of the documentary “Forever Peace Now”. His work is self-funded — no corporate sponsors, no government grants, no advertising.

Research, source verification, and structural editing assisted by Claude (Anthropic). Every statute, case, date, penalty figure, and quotation in this document was checked against primary or authoritative secondary sources. The arguments, conclusions, and moral claims are entirely those of the author.

© 2026 Vahid Razavi. Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International.

We Want to Hear From You!