Wait for the Catastrophe — That's the AI Plan

Responding to: AI: Full Speed Ahead in a Dark Unknown — Peggy Noonan · 2026-10-01

What the Piece Argues

An October 1, 2026 Wall Street Journal opinion column uses Bill Gates’s recent public warnings to argue that AI is the most dangerous technology humanity has ever approached, that nobody — not its creators, not its regulators — fully understands what it can do, and that the September 2025 White House self-policing agreement, drafted per Semafor by Meta’s Mark Zuckerberg, is the present arrangement. The single primary power-protecting talking point the piece advances is its fatalistic concession: “almost everyone involved in AI believes there will be at least one catastrophe, maybe more, likely including significant loss of life, before mankind and its governmental structures seriously and effectively attempt to regulate for safety.” This sentence converts an active policy choice (no binding safeguards) into inevitable weather, and the column’s closing observation of the industry executives’ silent body language at the White House ceremony as “a power move” reinforces the same move.

Receipts

The framing asks you to accept the AI industry’s voluntary self-policing as the deal we have, and to treat the coming catastrophe as weather rather than as a policy choice made by specific people against the public’s interest.

• Anchor: Semafor reported that Mark Zuckerberg’s side drafted the September 2025 White House AI agreement. The firms wrote the rule. The firms enforce the rule. The rule is a handshake.
• “More is being spent on speed than safety in the AI labs.” — this is the indictment; the column never names the policy lever that would change it.
• The September 2025 White House agreement is voluntary, drafted by the firms, enforced by the firms, and described by the President as “morally binding.” No third-party verification, no penalty for breach.
• The fatalism does the work: it converts an active policy choice (no binding safeguards) into an inevitable force of nature, so the reader stops looking for the people who made the choice.
• The “weird genius” frame launders unaccountability as exceptionalism. Calling the people running the most consequential technology in human history “weirdos” and treating that as a footnote, not as a reason to demand ordinary democratic accountability, is the rhetorical center of the piece.

What’s really going on:

• The concentrated beneficiaries are the firms’ executives, investors, and employees whose equity stakes rise with each quarter of unconstrained development; the diffuse cost-bearers are the public on whom the catastrophic risk is externalized. McKinsey Global Institute’s widely-cited June 2023 report “The Economic Potential of Generative AI” estimated generative AI’s potential annual value at up to $4.4 trillion across 63 use cases.
• Voluntary commitments have a documented breaking record: OpenAI’s board fired Sam Altman in November 2023 over safety concerns, employees and investors revolted, the board reversed inside five days, and the “superalignment” safety team was dissolved in May 2024 — coverage in The New York Times, Wired, and The Information.
• Industry-press reporting in 2025 placed OpenAI’s reported valuation near $500 billion and Anthropic’s near $183 billion, as reported, figures that reflect the upside of the present arrangement.
• The firms’ own chief executives have publicly acknowledged the catastrophic risk: Bill Gates (Meet the Press and a New York Times interview with Ezra Klein, both late September 2026), Sam Altman, Dario Amodei, and Demis Hassabis, in their own firms’ own communications.
• “Morally binding” is a phrase whose only enforcement is reputational; the firms invoking it have a documented record of treating reputation as a marketing channel rather than as a constraint.
• The European Union’s AI Act, formally adopted in 2024 and entering force in stages through 2026, classifies high-risk AI systems in healthcare, employment, and law enforcement, and binds vendors to conformity assessments, post-market monitoring, and meaningful fines — demonstrating that mandatory, enforceable regulation is operationally available when a legislature chooses to build it.

The Response Ladder

Polite Reframe

When to use: A persuadable moderate who has read the column and is sitting with the uneasy feeling that yes, this is dangerous, but also that nothing can be done.

Consider Maria, a forty-two-year-old ER nurse in Phoenix — a composite drawn from the kinds of cases already showing up in industry incident reports. Her hospital rolled out an AI sepsis-alert system last spring. In the vendor’s demo the system caught a septic patient six hours before crash. In Maria’s ER the system fires forty alerts a shift, the real ones disappear into the noise, and a sixty-four-year-old construction worker named Tomás walks in with a sore knee and goes home and dies of sepsis the next morning. Maria files a report. The vendor’s “morally binding” safety commitment says they will reduce false positives in the next release cycle. The hospital’s contract has a six-month optimization clause. Tomás is in the ground, and the next release cycle starts in three weeks.

This is the deal on offer. The companies writing the safety commitments are the companies whose products failed the patient. They write the rule. They enforce the rule. The rule is “we’ll get around to it.” And the column last week treated this arrangement with the rueful acceptance of someone describing weather.

There is a different deal available. The European Union’s AI Act classifies high-risk AI systems in healthcare, employment, and law enforcement and binds vendors to conformity assessments, post-market monitoring, and meaningful fines for non-compliance. It is not a perfect instrument. It is an instrument that, when a vendor’s system kills a patient in Phoenix, gives the hospital something other than a six-month optimization clause. The U.S. can build that instrument. Congress can build that instrument. State attorneys general, who are already suing AI vendors under existing consumer-protection and antidiscrimination law, are building parts of it without waiting.

The catastrophe is not inevitable. The catastrophe is the present arrangement. The present arrangement is voluntary. The present arrangement was drafted, per Semafor, by Mark Zuckerberg. “Morally binding” means the people whose system failed the patient get to decide when they have fixed it. The lever is the next election, the next state attorney general, the next member of Congress who can be reached at the local number, and the choice to stop treating this as weather and start treating it as a policy choice that specific people made and that specific people can unmake.

Mockery and Ridicule

When to use: The bystander at the Thanksgiving table who repeats the “yeah, but what can you do” line and needs to be shown the con.

Picture it. The Cabinet Room. September 2025. The President of the United States at the podium, beaming. Behind him, in a row, the men who own the most consequential technology in human history: Sundar Pichai, Satya Nadella, Mark Zuckerberg, Sam Altman, the man from Anthropic, the man from the new lab you have not yet heard of. They have agreed, the President announces, to act “safely and responsibly.” The agreement is, he says, “morally binding.” The men behind him are quiet, low-key, almost invisible. From the back of the room you would think they were furniture.

The men behind him, six weeks earlier, had their lawyers draft the agreement. Semafor reported this. The companies wrote the rule. The companies enforce the rule. The rule is a handshake. The handshake is “morally binding,” which in the language of corporate counsel means “voidable at the option of the agreeing party, with reputational consequences limited to a single news cycle.” The companies have a documented record of breaking their own public commitments when profit pressure arrives: OpenAI’s board fired Sam Altman in November 2023 over safety concerns, the employees revolted, the board caved inside five days, and the company’s “superalignment” safety team was dissolved in May 2024. The firms that cannot keep faith with their own boards are asking you to keep faith with their handshakes.

The President says the agreement is “morally binding.” He is correct. The moral in question is the one the firms are operating under. The binding is the binding on the people who are not in the room. The voluntary commitment is voluntary for the firms and mandatory for everyone the firms’ products touch. The voter who repeats the “what can you do” line is being read in.

You can do what every other regulated industry in American history has been told to do: accept that the product you sell is subject to mandatory safety standards enforced by the government, because the alternative is what we have right now, which is a man in San Francisco writing a rule and a man in Phoenix watching his patient die while the vendor’s lawyers run out the optimization clock.

Nuclear Satire

When to use: The reader who is past persuasion, who has heard the “morally binding” line and is looking for the language to match their fury.

They are building the second atomic bomb in real time. They are building it faster than the first one, with fewer controls, in more basements, with fewer physicists who will tell the newspapers what they have made, and the man at the podium says the agreement to control the bomb is “morally binding.” The agreement was drafted by one of the men building the bomb. The other men building the bomb stood behind the man at the podium and were quiet, low-key, almost invisible, and the column, looking at their body language, thought: “Huh, that’s a power move.”

The power move is the agreement. The power move is “morally binding.” The power move is the White House ceremony, six weeks after the firms’ lawyers drafted the text, with the President calling it a triumph and the men behind him calling it nothing, because they did not need to call it anything. The text was theirs. The enforcement is theirs. The “self-policing” is theirs. The “we’ll get around to it” is theirs. The “morally binding” is a phrase invented by people whose moral philosophy is the quarterly earnings call and whose binding is the contract lawyer’s narrow reading.

OpenAI’s board, in November 2023, did the one thing a self-policing industry structure was supposed to be capable of doing. The board looked at the conduct of the CEO and the safety team and decided the conduct could not stand. The board fired Sam Altman. The employees revolted. The investors revolted. The board caved inside five days. The CEO returned. The “superalignment” safety team was dissolved in May 2024. The board members who had voted to fire the CEO were, in time, replaced. The mechanism had been tested and the mechanism had failed. The mechanism had failed because the mechanism was inside the firm and the firm answered to investors.

The firms that cannot police a single board cannot police an industry. The firms that cannot keep faith with their own employees cannot keep faith with a White House handshake. The firms that dissolved their safety team when the safety team was inconvenient cannot be expected to honor a “morally binding” commitment when the commitment is inconvenient. The firms that drafted the agreement are the firms whose products are already failing in Maria’s ER in Phoenix, in Michigan’s MiDAS unemployment system (2013–2015 and aftermath), in the Dutch benefits scandal (the toeslagenaffaire / SyRI affair, ~2020) that destroyed tens of thousands of families. The firms whose products are killing people are asking you to accept their handshake as the deal.

The European Union built the AI Act anyway. The Act classifies high-risk AI systems in healthcare, employment, and law enforcement, binds vendors to conformity assessments and post-market monitoring, and provides for fines that exceed the cost of the optimization clause. The United States has not. The United States has, instead, the September 2025 voluntary agreement, drafted by Mark Zuckerberg, “morally binding,” enforceable by the firms that wrote it, against the people the products are failing. The catastrophe the column says is coming is not coming. The catastrophe is here. The catastrophe is the arrangement. The arrangement is voluntary. The arrangement was drafted by the men whose products are killing people, and the men at the podium were quiet, and the columnist thought it was a power move.

It is not a power move. It is a confession.

Profane Scorched-Earth

When to use: The reader who needs full catharsis and is done being polite about what the firms have done.

Let us speak plainly about what is happening in this country, because the language we have been given is built to obscure it, and the men doing the obscuring are rich enough to buy better language than the rest of us.

The firms are building a technology that their own chief executives, on television, in print, in private, say could kill a billion people. Bill Gates said it. Sam Altman has said it. Dario Amodei has said it. Demis Hassabis has said it. The firms’ own people are saying it. The men running the firms are saying it. The men who own the firms are saying it. They are saying it on the record, in the New York Times, in the Wall Street Journal, in the New Yorker, in congressional testimony, in podcasts, in essays, in blog posts, in the platform formerly known as Twitter. They are saying it in every venue that will have them. They are saying it and they are continuing. They are continuing because the money is on the other side of the catastrophe. The money is on the side of the speed.

The money is on the side of the speed because the speed is theirs. The risk is ours. The “morally binding” agreement is theirs. The drafting of the agreement was theirs — Semafor reported that Mark Zuckerberg’s people wrote the goddamn text. The enforcement of the agreement is theirs. The “self-policing” is theirs. The police are the criminals. The police are the criminals and they are being photographed at the White House and the President of the United States is calling the photograph a triumph and the column is calling the body language a power move. The body language is not a power move. The body language is the body language of a man who has just been handed a license and is pretending he has not been handed it because the camera is rolling. That is what a confession looks like when the man confessing can afford better lighting than the rest of us.

OpenAI’s board, in November 2023, did the one thing the voluntary structure was supposed to make fucking possible. The board fired Sam Altman. The board said: this man cannot be trusted with this technology. The employees revolted. The investors revolted. The board caved inside five fucking days. The CEO returned. The “superalignment” safety team — the team that was, in the firms’ own telling, the last internal mechanism for keeping the technology inside the human envelope — was dissolved in May 2024. The team is gone. The mechanism is gone. The firms’ own people, in the firms’ own public statements, are telling you the mechanism is gone, and the firms’ own public statements, in every other venue, are telling you the mechanism is “morally binding.” The firms are lying. The firms are lying in print. The firms are lying on camera. The firms are lying to Congress. The firms are lying to the President. The firms are lying to you.

The European Union built the AI Act anyway. The Act is a piece of legislation. The Act classifies high-risk AI systems and binds vendors to conformity assessments and post-market monitoring and fines that exceed the cost of the optimization clause. The Act is not perfect. The Act is an instrument. The Act is an instrument that, when a vendor’s system kills a patient in Phoenix, gives the patient’s family a remedy that the “morally binding” handshake does not give them. The United States has not built the fucking instrument. The United States has, instead, a voluntary agreement, drafted by the firms, “morally binding,” enforceable by the firms, and the men at the podium were quiet, and the column called the quiet a power move, and the quiet is the sound a man makes when he has been handed the keys to the city and is pretending he is still waiting in line.

The catastrophe is not coming. The catastrophe is the present arrangement. The present arrangement is voluntary. The present arrangement was drafted by the men whose products are killing people. The firms cannot police a single board. The firms cannot police a single safety team. The firms cannot police a single internal mechanism, and they dissolved the fucking mechanism when the mechanism was inconvenient, and the firms are asking you to accept their handshake as the deal. The firms are asking you to die quietly. The firms are asking you to die politely. The firms are asking you to die in the “morally binding” interval between one release cycle and the next. The firms are asking you to shut the fuck up and die on their schedule.

Do not die quietly. Do not die politely. Do not accept the handshake. The men who drafted the agreement are not weird geniuses. The men who drafted the agreement are men with a fiduciary duty to no one and a product that is failing in Maria’s ER in Phoenix, in Michigan’s broken MiDAS system, in the Dutch benefits scandal that wrecked tens of thousands of families. The men who drafted the agreement are men who dissolved their own safety team when the safety team was inconvenient. The men who drafted the agreement are men whose moral philosophy is the quarterly earnings call and whose binding is the contract lawyer’s narrow reading. The men who drafted the agreement are the police. The men who drafted the agreement are the criminals. The men who drafted the agreement are asking you to call this a power move so you do not have to call it what it is: a fucking confession.

Gates is right. The technology is the most dangerous thing humans have ever gone near. The technology is more dangerous than the atomic bomb because the atomic bomb required plutonium and a nation-state and a desert. The technology requires chips and wires and a data center and a man with a hoodie. The atomic bomb killed two hundred thousand people and ended a war. The technology, in the firms’ own telling, could kill a billion. The atomic bomb was regulated before it was deployed. The technology is being deployed and the regulation is being asked to wait. The regulation is being asked to wait because the firms wrote the rule that the regulation is being asked to wait for. The rule is “morally binding.” The binding is on us. The binding is on us and the men who wrote it are at the podium, and the men behind them are quiet, and the quiet is a confession, and the column called the quiet a power move because the column does not have the language to call a confession a confession.

The language is here. The language is this. The firms are the police and the firms are the criminals and the men at the podium know it and the men behind the podium know it and the column knows it and you know it. The only people who do not know it are the people who have agreed to die on the firms’ schedule. Do not be one of those people. The lever exists. The lever is the next election, the next state attorney general, the next member of Congress, the next refusal to accept a handshake as a substitute for a law. Pull the fucking lever.