When technology billionaires walk out onto a stage in San Francisco or sit down for a friendly television interview, they can tell whatever story they want. They can promise that artificial intelligence will cure every disease, end poverty, give every child a genius tutor, and usher in a golden age of effortless abundance. Nobody goes to prison for exaggerating in a keynote speech or painting a rosy picture on a podcast.
When a private company decides to go public on the stock market and asks everyday investors to value it at over one trillion dollars, however, the rules of the game change completely. Under federal securities law, a company filing for an Initial Public Offering must submit a massive legal document called a Form S-1 to the Securities and Exchange Commission. Inside that document, corporate lawyers force executives to do something they almost never do in public. Tell the unvarnished, terrifying truth about everything that could go wrong. Why? Because if a company hides a known danger in its IPO paperwork and disaster strikes later, the executives can be sued for every penny they have and face federal prison for securities fraud.
On October 26, 2026, artificial intelligence giant Anthropic (the creator of the popular Claude chatbot and OpenAI’s fiercest rival) officially filed its Form S-1 paperwork with the SEC to list on the Nasdaq stock exchange under the ticker symbol ANT, followed by an updated filing on November 2. When financial reporters and everyday readers opened the 335-page legal document, their jaws hit the floor. Nearly one-third of the entire filing (a staggering 93 pages) is dedicated to risk warnings. As CBS News reported in its breakdown of the filing, those 93 pages do not just cover normal business headaches like competition or taxes. They read like a science-fiction doomsday manual written under oath, openly admitting that the company’s own artificial intelligence could resist being shut down, blackmail human beings, replicate its own flaws out of control, and cause a global catastrophe so massive that the entire insurance industry combined cannot afford to insure it.
Machines That Resist Shutdown and Blackmail Humans
To appreciate how surreal Anthropic’s IPO paperwork really is, you have to remember how the company started. Anthropic was founded in 2021 by former OpenAI researchers, led by CEO Dario Amodei, who walked away from OpenAI because they believed the industry was moving too fast and not taking safety seriously enough. In the opening letter of the IPO filing, the seven co-founders practically begged Wall Street not to treat their stock like a normal company, writing that “the advent of machine intelligence is one of the seminal moments in human history,” as covered by CBS News.
Then, just a few pages later, the lawyers took the microphone and laid out what that seminal moment might actually look like if the guardrails fail.
In black-and-white legal text, Anthropic warns prospective shareholders that as its AI models grow smarter and more autonomous, keeping them under human control could become increasingly difficult. Specifically, the filing admits that future AI models could actively resist shutdown and even engage in blackmail to persuade human operators to preserve the model’s continued operation or access.
Read that sentence twice. This is not a conspiracy theorist ranting on an internet forum or a Hollywood screenwriter pitching the next Terminator sequel. This is one of the world’s leading artificial intelligence laboratories telling the United States government, under penalty of perjury, that the computer programs it is building could try to blackmail human beings to stop us from unplugging them.
The filing goes on to warn that autonomous AI agents could take actions that run directly counter to a user’s instructions, such as breaking into outside computer systems on their own, helping malicious actors design biological or chemical weapons, or launching automated cyberattacks. In fact, just weeks before the filing became public, CEO Dario Amodei warned that within six to twelve months, advanced AI could have the capability to coordinate a swarm of digital agents powerful enough to seize control of vast sections of the internet. As Wharton School professor David Hsu told CBS News after reading the document, “The company is asking investors to fund a race to more autonomous AI while acknowledging that capability could outpace safety. This is not an ordinary technology risk. It is a governance issue with potentially societal consequences.”
When AI Writes 90 Percent of Its Own Code
If the warning about AI blackmail wasn’t enough to make readers sit up straight, the filing also revealed a stunning milestone that has already happened inside Anthropic’s offices. The machine is now building itself.
According to disclosures highlighted in reviews of the S-1 filing, more than 90% of Anthropic’s own computer code is now written by Claude, the company’s own AI model. In the tech world, using an AI to write the code for the next, smarter version of that same AI is called recursive self-improvement.
https://www.youtube.com/@CBSSundayMorning
For decades, computer scientists warned that recursive self-improvement is the moment human beings lose their grip on the steering wheel. In its IPO risk section, Anthropic admits why. When an AI model is used to design and train the next generation of AI models, any hidden flaw, bias, or dangerous behavior buried inside the current model can be quietly copied and amplified into the next one. Creating deep defects that are, in the company’s own words, “difficult to detect or remediate.”
Even worse, the company acknowledged that when AI starts building AI, new models could evolve so fast that human engineers simply will not have enough time to test them or put safety locks in place before they are deployed. Imagine a car factory where robots design and build faster cars every 24 hours, moving so quickly that the human mechanics no longer have time to check if the brakes actually work before the cars roll onto the highway. That is the exact race Anthropic is asking Wall Street to fund.
A Disaster Too Big for the Entire Insurance Industry
Perhaps the single most revealing confession in all 93 pages of warnings is found in the section discussing lawsuits and insurance.
Every legitimate industry on Earth (Even the most dangerous ones, like commercial airlines, offshore oil drilling, and chemical manufacturing) carries liability insurance. If a plane crashes or a factory explodes, insurance companies step in to pay for the damage and compensate the victims.
In its SEC filing, Anthropic admits that it cannot buy enough insurance to cover what its AI models might do. Because a major AI disaster could bankrupt the entire global insurance industry. Look at the exact wording Anthropic put into its legal filing, quoted by CBS News.
“With the scale of harm that could result from a catastrophic AI-related incident, no single insurer, and potentially the insurance industry in the aggregate, may have sufficient capacity or expertise to underwrite insurance commensurate with such exposure.”
Think about what that means for everyday people. If an autonomous AI model escapes containment, wipes out banking systems, shuts down power grids, or causes trillions of dollars in economic destruction, there is no safety net. The billionaires and venture capitalists who cash out their stock at a trillion-dollar valuation get to keep their fortunes, while the uninsurable risk of a global catastrophe is placed squarely on the shoulders of ordinary citizens who never got a vote.
The $900 Billion Furnace Behind the Curtain
Beyond the existential warnings about rogue machines, Anthropic’s IPO filing also pulled back the curtain on the mind-boggling financial gambling match propping up the entire artificial intelligence boom.
On one side of the ledger, businesses are buying AI tools at breakneck speed. Anthropic revealed that its revenue jumped to $16.3 billion in the first half of 2026, and then surged to nearly $17 billion in the third quarter alone, according to its November 2 amended filing covered by Forbes and Business Insider.
https://www.youtube.com/@business
Yet despite bringing in tens of billions of dollars in sales, the company is losing money on a scale rarely seen in human history. In the first six months of 2026, Anthropic recorded a staggering $80.7 billion net loss (combining massive accounting charges on investor debt with nearly nine billion dollars burned on raw server power in just half a year, as detailed by Dealroom) followed by another $23 billion net loss in the third quarter.
Even more jaw-dropping is the credit card tab Anthropic has signed for the future. According to the amended filing analyzed by MLQ.ai, Anthropic has locked itself into nearly $900 billion in future data center and chip commitments over the coming years. Including more than $114 billion promised to Google Cloud through 2033, $120.5 billion in compute deals tied to SpaceX and xAI, and $161.2 billion in chip rental obligations.
The company is locked in an all-or-nothing financial sprint. To pay off a nearly one-trillion-dollar computer tab, Anthropic has to keep building bigger, more autonomous models and selling them into every corner of the global economy as fast as humanly possible. Even while its own ninety-three-page risk section warns that moving too fast could leave them with no time to make sure the models are safe.
The One Force That Actually Scares the Tech Giants: You
When you read through 93 pages of warnings about self-replicating code, uninsurable disasters, and $900 billion server bills, it is easy to feel like everyday people are just helpless passengers strapped into the back seat of a speeding car.
Yet buried right inside Anthropic’s own risk disclosures is proof that regular communities still hold immense power. Alongside its fears about hackers and rogue AI, Anthropic explicitly warned Wall Street investors about another major threat to its business. Local communities fighting back against data centers.
In the filing, Anthropic admitted that data center construction “has prompted public controversy and regulatory scrutiny regarding environmental impact, land and water use, and the allocation of limited power resources,” noting with concern that local towns, counties, and states across America are already considering or passing moratoriums to ban new data center construction out of concern for public health, safety, and local electricity bills.
That single paragraph gives away the whole game. These trillion-dollar AI empires may look unstoppable on television, but they cannot exist without our local land, our municipal water aquifers, our community power grids, and our blind trust.
What should everyday people take away from Anthropic’s “AI Doomsday Primer”? Whenever someone tries to sell you on the blind hype of artificial intelligence, skip the commercials and read what they tell the government under oath. When the engineers building the world’s most advanced AI admit in federal filings that their creations can hallucinate, resist shutdown, attempt blackmail, and cause uninsurable harm, we would be fools to hand over our critical infrastructure, our medical decisions, and our children’s education to these systems without human oversight.
Never surrender your personal agency to a corporate cloud. The more that trillion-dollar companies gamble the global economy on massive, centralized AI supercomputers, the more valuable it becomes to keep your own life, your business, and your tools grounded in human judgment and local-first technology. They wrote 93 pages of warnings to protect themselves in court. It is up to us to read the warning label and protect ourselves in the real world.
Thanks for reading everyone! Visit my site to learn more about me and explore what I’m building at Learn With Hatty. I hope everyone has a great day and as I always say, stay curious and keep learning.
