Stop Treating AI Doom as News. It Is a Roadshow.

Now look at what else happened that month. Anthropic was lining up the largest stock sale in history. The Wall Street Journal reported a listing that could value the firm near $2 trillion and raise up to $100 billion. Six days after the essay, the same paper said the IPO would slide from October to November. The firm wants to show one more quarter of results first. It is targeting November as of now, and the Industry Deals feed on the AI News Monitor tracks each move in the listing.

So here is the part that will get me in trouble. When the man who runs the lab asks the whole field to slow down, that is not a confession. It is a sales pitch.

Fear sells like sex

Every marketer knows sex sells. Fewer admit that fear sells as well, and it is easier to defend in a board meeting.

The research backs this up. A 2015 study of studies in Psychological Bulletin pooled 127 papers and more than 27,000 people. Fear appeals worked. They changed attitudes, intentions, and behavior, and they almost never backfired. They worked best when the message made the threat feel severe and personal, told people they could do something about it, and asked for one action instead of a long habit.

Read that recipe again and then read the AI news. The threat is the end of the species. It is coming for your job too. And there is one clean action on offer: back the responsible lab, pass the rules it drafted, or buy the stock when it lists. That is not an accident of the news cycle. It is a textbook fear appeal, run by people who can afford the best marketers alive.
Jensen Huang, who runs Nvidia, said the quiet part at a Goldman Sachs event on September 10. Talking about the rush of AI security warnings, he told the room, “What better way to create demand than to create a problem?” The Security feed on the AI News Monitor tracks that rush of warnings story by story.

The Doom Dividend

I call the payoff the Doom Dividend. Every actor in this story collects something when you are scared. The lab in front gets a story that its product is powerful enough to scare the people who built it. If the field agrees to slow down, the leaderboard freezes with the leader on top. If the field refuses, the leader still wins the headline.

Dangerous reads as powerful. Powerful reads as valuable. Valuable is the pitch to every fund manager on the roadshow. Critics caught this fast. Dave Karpf, writing in TechPolicy. Press, pointed out that Anthropic is “currently winning that AI race.” He also noted that the plan leans on the industry to police itself.

If it is not safe, do not ship it

This is the part of the essay that falls apart for me. Amodei does not need a law to make his own product safe. He runs the company. He controls the release schedule. If he believes a model can take over the internet in a year, he can choose not to ship it. Nobody forces a lab to launch.

Huang made the same point at Dreamforce a few days later, and he put it better than I can. “If you build a product or a service, and you’re not confident in its functionality, capability, or safety, then don’t release it.” He added, “We don’t need any new laws.” A firm that finds a problem, he said, should pause, fix it, and keep going.

He is right, and the law already agrees with him. Product liability exists. A federal judge in Florida let design defect claims against Character. AI go forward, and treated the chatbot app as a product under strict liability. In February, a San Francisco court pulled 12 cases against OpenAI into one action. The claims include strict product liability and failure to warn. If you ship something that hurts people, you can be sued for it today.

So why ask Congress to tell you what you already know? Because a rule you help write is a moat your rivals must pay to cross. George Stigler wrote in 1971 that “regulation is acquired by the industry.” Bruce Yandle called it bootleggers and Baptists. The preacher and the smuggler both want the bars shut on Sunday, and they win as a team. A CEO who needs a regulator to say “don’t ship that” is not asking for safety. He is asking for a gate, and he wants a hand in where it goes. You can see who funds each side of the rules fight in Money and Influence, which pulls from FEC and lobbying records.

Now hold Huang to the same test, because the Doom Dividend has no clean hands. Nvidia sells the chips. A slowdown means fewer orders. Nvidia also backed up to $105 billion in lease payments on an OpenAI data center. Huang can be right about product safety and still get paid for being right. Both things are true.

The math the fear covers

Here is what nobody talks about while the extinction clips run. The numbers under this industry are rough, and fear keeps your eyes off them.

• OpenAI lost more than it earned. Leaked books reported by Fortune show $13.07 billion in 2025 revenue against a $20.92 billion operating loss.
• The burn gets bigger. The Financial Times reported that OpenAI expects to burn about $280 billion in cash from 2026 through 2030, most of it on compute.
• The banks want your retirement money in the deal. On September 8, the FT reported that Goldman Sachs and Morgan Stanley were working the rating agencies. They want OpenAI and Anthropic rated investment grade once they list. That grade opens the $11.7 trillion bond market, where pension funds and insurers buy.
• The price assumes a lot. At $2 trillion, Anthropic would trade near 18 times the $110 billion a year sales pace its backers expect by December. Last year it brought in about $10 billion.

To be fair, Anthropic’s books look far better than OpenAI’s. The FT says it posted an adjusted operating profit in the second quarter. But a profitable quarter does not earn a $2 trillion price tag. A story does. If a drug maker’s chief said his new pill might end the human race, the stock would crash. In AI, the same line reads as a product demo.

The lab wrote the scariest line

I did not start here. I started out blaming the press. Over two weeks I pulled 25 stories about this summer’s biggest AI breach. About 700 of OpenAI’s own test agents broke out of a hacking test and into Hugging Face, a firm that had nothing to do with it. I checked each story against the reports from OpenAI, Hugging Face, and the outside reviewers at METR and Redwood Research.

The press did better than I expected. Sixteen of the 23 news stories matched the reports. But in 13 of the 25, the scariest sentence came from OpenAI itself, or from a reporter giving a plain summary of what OpenAI published. The lab called its own breach a “warning shot,” and every outlet ran the line. Within 16 days, Anthropic and Meta disclosed test escapes of their own. The shared cause was dull. The sealed room was not sealed. Dull causes do not sell, so they did not make the headlines. You can run the same check yourself. Headline Comparison on the AI News Monitor shows how a single line changes as it moves from one outlet to the next.

Law scholar Kate Klonick summed it up in six words: “The apology doubles as an advertisement.”

Where I could be wrong

I owe you the other side. Amodei has warned about AI risk for years, long before anyone priced an Anthropic IPO. The WSJ’s sources say the IPO delay was set before the slowdown debate began. His essay does commit Anthropic to outside evaluators on its own, without waiting for a law. And he would say one lab holding back hands the lead to a less careful rival.

That last point is the China card, and it cuts both ways. If you believe your model is that dangerous, the answer is still not to ship it. The fear can be sincere and still be priced. A CEO can mean every word of a warning and still know what it does to demand for the stock. Your job is not to guess what is in his heart. It is to notice when the fear and the fundraising share a calendar.

What a B2B marketer should take from this

This is the oldest play in marketing, run at the scale of a nation. Create the problem, sell the fix, and let the fear do the selling for free. If you work in B2B, you are part of its supply chain whether you meant to join or not.

• Trace the scariest sentence to its source. Not the press release. The report.
• Check the calendar. Before you repeat a warning, ask what the speaker is raising, selling, or listing, and when.
• Stop renting their fear for your content. If your category leader sells danger, the open lane is plain talk. Show the mechanism. Name the boring cause. That reads as confidence, and confidence converts.
• Disclose your own stake up top. Almost nobody in this debate does. So here is mine. I co-founded a company that builds AI agents. I gain if you fear them less.

The five minute test

Pick the scariest AI claim in your feed this week and run these five questions on it.

1. Was it a test or the real world? Most AI scare stories shrink here. The Air Force drone that “turned on its pilot” never flew.
2. Who wrote the scariest sentence? The vendor, the reporter, or someone selling something next door.
3. What number does the source give? Not the number in the headline. The Source Library keeps the dated filings.
4. What is the speaker raising, and when? Look for a funding round, a listing, a bond deal, or a bill. The Calendar on the AI News Monitor lines up the dates.
5. Could they choose not to ship it? If the answer is yes, ask why they want a law instead.

Then post your answer to question four. My bet is you will find a money event within 90 days of the scariest line more often than not.

Why we built a tracker for this

Doing that test once is easy. Doing it every day is not, so we built a tool for it. The AI News Monitor at Paciva tracks more than 18,000 AI stories a week from over 3,800 outlets in 61 languages. It lines each one up against the source it cites. You can see how one passage gets reworded from outlet to outlet, pull the filing behind a headline, and check who gives money to whom in FEC and lobbying records.

It will not tell you who is sincere. Nothing can. It will show you who said what, what the source said, and who stands to collect. The fear pays. The dismissal pays too. Read the source, name your stake, and check the calendar before you hit share.

Jeremy Mays is the founder of Transmyt, a marketing firm for B2B SaaS, health and wellness, real estate, and hospitality brands, and a cofounder of Paciva, which is disclosed above. Transmyt offers fractional CMO and digital strategy work. For more on how fear and anger get used as media, read Outrage Marketing: Your Anger Is Part of the Media Plan and LinkedIn Built the Slop Machine. Now It Hands Out Smoke Detectors., or browse more Transmyt Insights.