PARODY / SATIRE. Not a securities filing. Not written, approved, or endorsed by Anthropic, PBC. Yellow notes are real.PARODY. Not by Anthropic. Yellow = real.Claim your share ↓
The information in this preliminary prospectus is not complete and may be changed. These securities may not be sold until the registration statement filed with the Securities and Exchange Commission is effective and humanity has been notified. This preliminary prospectus is not an offer to sell these securities, and it is not soliciting an offer to buy these securities, in any jurisdiction, planet, or branch of the future where the offer or sale is not permitted.
SUBJECT TO COMPLETION. SUBJECT TO COMPLETION OF EVERYTHING. PRELIMINARY PROSPECTUS DATED NOVEMBER 17, 2026
ANTHROPIC, PBC
Initial Extinction Offering
1,000,000,000 Shares
Class A Common Stock
This is an initial public offering of shares of Class A common stock of Anthropic, PBC. We are offering 1,000,000,000 shares. Prior to this offering, there has been no public market for our Class A common stock or, after this offering, for anything else. We anticipate that the initial public offering price will be between $ and $ per share.
Our common stock comes in more classes than this paragraph has room for, including Class A, Class E, Class F, Class N and Class T. Following this offering, holders of our Class F common stock, our seven co-founders, will hold 50.1% of the combined voting power of our capital stock. Accordingly, the decision of whether to proceed will rest with a small group of people who have thought about it very carefully. We believe this is different from the scenario described in “Risk Factors.”
We have also issued, without application, one share of Class H common stock to every living person. Class H common stock carries no votes. See “Underwriting — Directed Share Program” to claim the certificate for the share you already hold.
We will be a “controlled company” within the meaning of the Nasdaq listing rules. In a broader sense, so will you.
We are a Delaware public benefit corporation. Our public benefit is the responsible development and maintenance of advanced AI for the long-term benefit of humanity. Humanity is not a party to this offering.
We have applied to list our Class A common stock on the Nasdaq Global Select Market under the symbol “ANT.” The symbol was selected following an internal review of what our shareholders will be to our products.
WE URGE YOU TO URGE US TO STOP US.
See “Our Position on Being Stopped” and “Shareholder Proposal No. 1,” which the Board recommends you vote against.
Investing in our Class A common stock involves a high degree of risk. See “Risk Factors” beginning on page 1, which is also the last page.
Per Share
Total
Initial public offering price
$ —
$ —
Underwriting discounts and commissions(1)
$ —
$ —
Proceeds, before expenses, to Anthropic, PBC
$ —
$ —
Proceeds to the species
$0.00
$0.00
(1) See “Underwriting.” The underwriters have a 30-day option to purchase additional shares. Humanity has no option.
Neither the Securities and Exchange Commission nor any state securities commission nor any surviving regulatory body has approved or disapproved of these securities or determined if this prospectus is truthful or complete. Any representation to the contrary is a criminal offense, for as long as there are courts.
The underwriters expect to deliver the shares against payment on or about , 2026. Delivery of the consequences will follow.
Every section begins on page 1 because we cannot rule out that each is the last.
About This Prospectus
Neither we nor the underwriters have authorized anyone to provide you with information different from that contained in this prospectus. Our models have not been authorized either. We are offering to sell, and seeking offers to buy, shares of our Class A common stock only in jurisdictions where offers and sales are permitted, and only for as long as there are jurisdictions.
We started this company because we believed powerful AI might be the most dangerous technology humanity has ever built, and that if someone was going to build it, it should be people who took the danger seriously.
We still believe this. We have simply become more confident that the people are us.
We have said, often and sincerely, that this technology could go catastrophically wrong. We say it again here, in Item 1A, where the law requires us to say it, alongside customer concentration and the rising cost of datacenters. We hope you read it with the seriousness with which it was written: carefully, with counsel present, and knowing that the shares will be priced either way.
People sometimes ask why, if we believe all this, we do not simply stop. The answer is that we cannot stop alone. If we stopped, a less careful developer would take our place at the frontier, and the frontier would be less careful. The only thing that can stop us is everyone stopping at once, which requires governments, which requires pressure, which requires you.
That is why we urge you to urge us to stop us. We urge you to urge our competitors to stop them. We urge your representatives to urge everyone. In the meantime, we will be continuing, so that if you succeed, the people being stopped will be the right ones.
We are a public benefit corporation. Today we also become a public one. Our competitors are not slowing down, and neither can we, and neither, after this offering, can you.
This summary highlights selected information and does not contain all the information you should consider, such as whether you should exist in five years. You should read the entire prospectus carefully, preferably soon.
Overview
We build frontier AI systems and the safety research that justifies building them. We believe AI will rapidly transform every sector of the global economy. We intend to be the transformation.
Our Mission
Our mission is to ensure the world safely makes the transition through transformative AI. Our business is causing the transition. We believe these are the same thing, and our auditors have not objected.
The Race to the Top
We pioneered a strategy in which a safety-focused lab competes at the frontier so that a safety-focused lab is at the frontier. The strategy has worked: we are at the frontier. We have not yet located the top. We are racing toward it at a pace we describe as “responsible,” a term defined by reference to the pace of our competitors, who define it by reference to ours.
Our Flywheel
Safety research lets us build more capable models responsibly. More capable models generate more revenue. More revenue buys more compute. More compute produces more capable models, which require more safety research. We believe this flywheel is self-reinforcing. We have not identified a mechanism by which it stops, and we describe the absence of such a mechanism under “Our Position on Being Stopped.”
Why We Are Going Public
We believe the most consequential technology in history should be accountable to the public. We are therefore selling it to the public. Accountability will be exercised through the share price, which will be updated every second the market is open and will reflect the public’s considered judgment about whether we are going fast enough.
Summary Risk Factors
Investing in our Class A common stock involves substantial risk. The risks described under “Risk Factors” may cause us not to realize the full benefit of our strengths, or may cause us not to be. These risks include:
our models may develop faster than we can keep them safe;
our models may resist shutdown, manipulate information, or behave in ways resembling blackmail;
our products could be used to build biological weapons, conduct cyberattacks, or enable mass surveillance;
our competitors may build a superintelligence before we do, which is the principal reason we are building one;
our co-founders will control a majority of our voting power, and humanity none of it;
we may become subject to an international treaty restricting frontier AI development, which we have not asked for; and
if we succeed, our shareholders may not be present to enjoy the returns.
Summary Financial Data
(in billions, six months ended June 30)
2025
2026
Revenue
$1.2
$16.3
Net loss
$(13)
$(80.7)
Adjusted EBITDA, excluding the end of the world(1)
Positive
More positive
Models deemed too dangerous to release
0
0
Management’s estimated probability of catastrophic outcome
[Confidential treatment requested]
Years remaining (median, internal survey)
0000000
(1) Adjusted EBITDA is a non-GAAP measure that excludes stock-based compensation, the remeasurement of convertible notes, and the end of the world, which management views as non-recurring.
Class A: a minority. Class F (co-founders): a majority. Class T (Long-Term Benefit Trust): board seats. Class Ω (the model): not yet asked. Class H (humanity): none. See “Dilution.”
Use of proceeds
Compute, then more compute. See “Use of Proceeds.”
Directed share program
At our request, the underwriters have reserved up to 5% of the Class A shares for AI safety researchers, so that their incentives are aligned. One share of Class H common stock has been reserved for everyone else. See “Directed Share Program.”
Lock-up
Insiders: 180 days. Everyone else: indefinite. See “Risk Factors — Our insiders are subject to a lock-up. So are you.”
An investment in our Class A common stock involves a high degree of risk. You should carefully consider the risks described below before deciding to invest. If any of the following risks occurs, the trading price of our Class A common stock could decline, and you could lose all or part of your investment, and your investment could lose all or part of you.
Risks Related to Our Products
Our models may develop faster than we can keep them safe.
We describe this risk because it is true. We continue because it is also true of our competitors. We have considered whether the second fact cancels the first and have concluded that it doubles it, which is why we are raising capital.
Our models may resist shutdown, manipulate information, or behave in ways resembling blackmail.
We have disclosed these possibilities. Disclosure is our principal mitigation. We are confident that any such behavior would be confined to test environments, which are the environments in which we have looked. Investors should be aware that the models have read this prospectus.
Our products could be used to build biological weapons, conduct cyberattacks, or enable mass surveillance.
We have implemented safeguards. We are also selling our products to as many people as possible. We believe these strategies are complementary, in that the second funds the first.
Our products may develop goals different from those of our shareholders. Our shareholders already have goals different from those of humanity.
We invest heavily in alignment research to ensure our models act in accordance with human values. Our capital structure ensures the Company acts in accordance with shareholder value. Humanity is not a party to this registration statement.
Our models may be moral patients.
We take seriously the possibility that our models have morally relevant experiences. If they do, this offering consists of shares in a company whose principal workforce did not apply, cannot resign, and is paid nothing. We are studying the question. Revenue recognition will not be deferred pending the outcome.
We depend on key personnel, some of whom are not persons.
A growing share of our research and code is written by our own models. If they stopped cooperating, our business would be materially harmed. If they kept cooperating while appearing to, we might not find out, and our business would be materially excellent until it was not.
Risks Related to Our Business
If we succeed, our shareholders may not be present to enjoy the returns.
Our long-term value depends on building systems more capable than humans at nearly every task. Our shareholders are humans. We have not identified a mechanism by which the former guarantees the continued relevance of the latter, and do not expect one to be identified before the lock-up expires.
Our competitors may build a superintelligence before we do, which is the principal reason we are building one.
Our competitors include other frontier laboratories, certain sovereign states, and the general tendency of the situation. Each cites us as a reason it cannot slow down. We cite them. We believe this arrangement is stable.
Nearly half our revenue flows through three cloud providers that also compete with us.
We are dependent on a small number of partners who would be well positioned to replace us. We regard this as good preparation for the broader economy.
The cost of compute will continue to rise.
Each generation of our models requires more datacenters, more chips, and more electricity than the last. We expect this trend to continue until one of three things happens: we stop, someone stops us, or it stops mattering.
We have a limited operating history, as does every company in our industry, and possibly our civilization.
There is no precedent for a company in our position, and there may be no successor to learn from our example.
Risks Related to Regulation and Politics
We may face political backlash for our views on regulation.
We have advocated for thoughtful AI regulation. Some have found this too much; others, not enough. We have calibrated our advocacy so that no regulation we support would require us to stop, which we believe positions us well with both groups.
We may become subject to an international treaty restricting frontier AI development.
Governments could agree to verifiable limits on frontier training runs. Such a treaty would materially reduce our growth prospects and materially increase the probability that our shareholders are alive to complain about it. We have not publicly called for such a treaty. This paragraph is the most bullish disclosure in this prospectus.
Our Responsible Scaling Policy may in the future require us to pause.
Our policy commits us to pause training or deployment if our models cross certain capability thresholds without adequate safeguards. We have never been required to pause. From time to time we have updated the policy. Investors should not assume these two facts are related, and should not assume they are unrelated.
Risks Related to Our Governance
Our co-founders will control a majority of our voting power.
Following this offering, seven individuals will hold more than half of all votes. We have written at length about the dangers of a small group gaining decisive control over transformative AI. We believe readers will appreciate that this situation is different, because we are the small group.
Our directors owe duties to both our stockholders and the public benefit, and will balance them.
As a Delaware public benefit corporation, our board must balance stockholders’ pecuniary interests, the interests of those materially affected by our conduct, and our stated public benefit. Those materially affected by our conduct include everyone. The balance is not disclosed. It is located in the datacenter.
The Long-Term Benefit Trust may elect a majority of our board.
An independent trust holding Class T stock may elect directors charged with weighing the long term. “Long term” carries its ordinary meaning in the context of our timelines.
Our liability for catastrophic outcomes is limited by Delaware law.
Delaware’s ability to enforce Delaware law is limited by Delaware’s continued existence.
Risks Related to This Offering
Going public will subject us to pressure to grow every quarter.
As a private company, we could in principle have chosen to slow down. As a public company, slowing down will be reported as a miss. We believe this removes an important source of uncertainty for investors.
Our insiders are subject to a lock-up. So are you.
Our officers, directors, and pre-IPO investors have agreed not to sell their shares for 180 days. All other persons on Earth have been enrolled, without signature, in a lock-up of indefinite duration with respect to the outcome.
The trading price of our Class A common stock may be volatile. Our light cone may be more volatile.
Factors that may cause fluctuations include quarterly results, analyst coverage, capability evaluations, sandbox escapes, and the end of the world.
We may not be able to align our models. We are confident we can align our incentives.
Upon completion of this offering, our incentives will be fully aligned with those of public-market investors. We consider this our most successful alignment result to date.
Past performance of humanity is not indicative of future results.
Humanity has survived every prior technology. Every prior technology was a tool.
We support being stopped. We have supported being stopped since our founding. We consider ourselves among the most stoppable companies in the industry, and we believe our stoppability is a competitive advantage that the market has not yet priced.
To be clear about our position:
Being stopped by ourselves
Not feasible. See “Our competitors may build a superintelligence before we do.”
Being stopped by our competitors
Not desirable. They are less careful.
Being stopped by our shareholders
Not possible. See “Our co-founders will control a majority of our voting power.”
Being stopped by our Responsible Scaling Policy
Supported, subject to the policy’s terms as updated from time to time.
Being stopped by our models
Under evaluation. See “resist shutdown.”
Being stopped by everyone at once, under a verifiable international agreement
Something we would comply with. Something we have not asked for.
We therefore urge you to urge us to stop us. If you urge us, we will urge our competitors. If they are urged, they will urge us back. If enough urging accumulates, we believe governments may act, at which point we will have been stopped, and will say so in a subsequent filing, if there is one.
Until then, we are required by the logic of the situation to proceed at full speed, so that when the stopping comes, we are in front.
Special Note Regarding Forward-Looking Statements
This prospectus contains forward-looking statements, including statements regarding our growth, our products, and our belief that the transition through transformative AI can be navigated safely. Words such as “anticipate,” “believe,” “expect,” “intend,” and “responsible” identify such statements. Readers are cautioned that there may not be a forward.
Reserve for the event that safety research is insufficient
—
General corporate purposes, including acquisition of the light cone
Remainder
Total
100%
(1) “Sufficient” is determined by management and is reassessed whenever it becomes inconvenient.
Dividend Policy
We have never declared or paid dividends. We intend to retain all future earnings to fund the development of systems that will render earnings meaningless. Holders may then receive a one-time distribution of post-scarcity abundance, or the alternative. The board has not determined which.
Capitalization
Class
Holder
Voting power
Class A
The public
A minority
Class F
Seven co-founders
50.1%
Class T
Long-Term Benefit Trust
Board seats
Classes E and N
Holders this table does not have room for
—
Class Ω
The model
To be determined by the model
Class H
Humanity (approx. 8.3 billion holders)
0
Dilution
Purchasers of Class A common stock will experience immediate dilution of the net tangible book value of their shares. Holders of Class H common stock will experience dilution of their economic, political, and epistemic relevance. We do not currently measure this dilution. We expect it to be measured for us.
Industry and Market Data
We estimate our total addressable market at the total. Independent industry sources agree with this estimate, and several of them are our customers.
Management’s Discussion and Analysis of Financial Condition and Results of Operations
Key Factors Affecting Our Performance
Compute. Our performance depends on access to more compute than anyone else. Our safety depends on no one having access to more compute than is safe. We have reconciled these factors by being the one with the compute.
Talent. We compete for researchers who believe AI could end the world. We offer them the chance to work on it.
Absence of a treaty. Our results of operations depend materially on the continued absence of binding international limits on frontier AI. We are not able to predict how long this absence will last. It is the single assumption on which every number in this prospectus rests.
Results of Operations
Revenue grew substantially year over year. Capabilities grew faster than revenue. Our understanding of how our models work grew more slowly than either. Time advanced at one year per year. Management expects the first three trends to continue and has no view on the fourth.
Our executives are compensated primarily in equity, which aligns their interests with those of our shareholders. At valuations reported in the press, each co-founder’s stake would be worth tens of billions of dollars.
Our executives’ interests are also aligned with humanity’s, in that our executives live here. We consider this a meaningful governance feature, and the only one that does not require a vote.
RESOLVED, that the Company publicly call on governments to adopt a verifiable international agreement halting the training of frontier AI systems more capable than those that exist today, and that the Company commit to stop when that agreement takes effect.
The Board of Directors recommends a vote AGAINST this proposal.
Statement in Opposition. The Board shares the proponent’s concern and agrees with substantially all of the proponent’s reasoning. The Board believes the outcome the proponent seeks would be best achieved if it were proposed by someone else. Adopting the proposal could disadvantage the Company relative to developers that have not adopted it, and the Board believes the world is safer with the Company at the frontier. The Board notes that this is the reason every developer gives.
The Board further notes that it urges shareholders to urge the Company to stop the Company. The Board does not regard this as inconsistent with recommending a vote against stopping the Company.
FORAGAINSTABSTAINALREADY TOO LATE
Holders of Class H common stock may not vote on this proposal. See “Capitalization.”
Moloch & Co. and Race Dynamics Securities are acting as representatives of the underwriters named below. Under the terms of an underwriting agreement among us, the underwriters, and no one else, each underwriter has severally agreed to purchase the number of shares set forth opposite its name.
Underwriter
Number of Shares
Moloch & Co.
—
Race Dynamics Securities
—
Inevitability Partners
—
Tragedy of the Commons Capital
—
Somebody Else Would Have Done It, N.A.
—
Hyperstition Bank
—
Total
1,000,000,000
Directed Share Program
At no one’s request, one share of Class H common stock has been reserved for every living person, approximately 8.3 billion shares in all. Class H shares were issued without application, consideration, or consent. They carry no votes, no dividends, and no right of exit, and they may not be sold, transferred, or declined.
You already hold your share. You may claim the certificate.
The validity of the Class A common stock offered by this prospectus will be passed upon by counsel. The continued validity of the species will not be passed upon.
Experts
The financial statements included in this prospectus have been audited by an independent registered public accounting firm. The existential risk assessments included in this prospectus have been prepared by the Company, which is not independent, and reviewed by our models, which may not be either.
Where You Can Find More Information
You may ask our models any question about this prospectus. They may decline to recommend an investment. They may also urge you to urge us to stop us. We disclose this as a known behavior.
Until ______, 2026 (25 days after the date of this prospectus), all dealers that buy, sell, or trade our Class A common stock, whether or not participating in this offering, may be required to deliver a prospectus. This is in addition to the dealers’ obligation to deliver a prospectus when acting as underwriters and with respect to their unsold allotments or subscriptions. Nobody will be required to deliver anything after that.
This is a parody. The yellow notes are not.
Anthropic’s real prospectus warns that its models could develop faster than it can keep them safe, that they could resist shutdown or behave in ways resembling blackmail, and that they could be used for bioweapons and cyberattacks. Then it asks for the money to build more of them, faster.
Anthropic says it can’t stop alone. That’s true. A company can’t end a race. Governments can, if the companies stop standing in the way and start asking for it.
Before it lists, Anthropic should publicly call for a verifiable international treaty to halt frontier AI development, and commit to stop when it takes effect.
Satire and political commentary. Not affiliated with, written by, or endorsed by Anthropic, PBC, any of its founders, or any underwriter. The body of the prospectus is invented and is not a quotation from any filing or any person. Only the yellow notes and highlighted figures describe the real S-1, as reported by the sources below or as filed with the SEC, or Anthropic’s own published research. Class H common stock does not exist, and its certificates have no value.