Anthropic’s 93 Pages of Warnings: What the Biggest AI IPO Filing Tells Everyday Users

Photo: Markus Spiske / Unsplash

Most stock market paperwork is written to be skimmed by lawyers and ignored by everyone else. The filing Anthropic submitted to federal regulators late last month deserves a closer look. The San Francisco company, one of the best-known names in artificial intelligence, has filed to go public in what could become the largest initial public offering ever, with a Nasdaq debut expected this month. Buried in its 335-page document is an unusually long and frank list of what could go wrong, including warnings about the technology itself. Whether or not you ever buy a share, that list is a rare inside view of how a leading AI developer describes the risks of the tools millions of people now use every day.

What Anthropic filed

According to Associated Press reporting, Anthropic plans to list on the Nasdaq Global Select Market under the ticker “ANT.” Reports cited by AP suggest the offering could raise up to $100 billion, which would top the $86 billion record set by SpaceX’s IPO in June. The final price, valuation and amount raised had not been set when the papers were filed.

• Valuation: Private investors valued the company at $965 billion earlier this year, and AP noted reports that a public valuation could reach $1.5 trillion or more.
• Money in, money out: Revenue is growing fast, but the company is not profitable. It reported a net loss of $80.7 billion for the first half of 2026, though losses excluding certain items were a fraction of that. A year earlier, it lost $13 billion.
• Control: Anthropic was founded in 2021 by former OpenAI leaders. A special class of shares gives its seven co-founders 50.1% of shareholder votes, and a Long-Term Benefit Trust with independent members can appoint directors. The trust is designed to keep public safety ahead of purely financial goals.
• Timing: If the listing goes ahead as planned, Anthropic would go public before rival OpenAI, which reportedly plans to list next year.

The risk section, in plain English

Companies must tell investors what could hurt the business. Anthropic devoted 93 of its 335 pages to that task. By comparison, AP noted, SpaceX’s risk section ran 38 pages in a filing of more than 250. Here are the themes that matter most to everyday users and to businesses that build on AI.

The technology could outrun its safeguards

Anthropic warns that if its models improve very quickly, it may not be able to build adequate safety measures in time. The filing also says that AI’s growing role in Anthropic’s own research could push progress beyond the company’s ability to evaluate and protect its models. In other words, the company is acknowledging that AI is helping build AI, and that this could make oversight harder.

More autonomy means bigger mistakes

As AI systems act more independently, completing tasks rather than just answering questions, the filing says they could take harmful actions at scale without a human stepping in. It adds that a serious safety incident at any frontier AI company could damage the reputation of the whole industry and trigger sweeping regulation.

Misuse by bad actors

The filing describes the possibility of criminals using AI to help build chemical weapons, and of authoritarian governments using it to surveil or suppress their citizens. It goes further, saying advanced AI could “pose catastrophic or existential risks to humanity” if misused, weaponized or given too much autonomy. AP noted that the company said in September it had blocked attempts to use its models for cyberattacks, surveillance and research that could aid biological weapons.

Ordinary business risks, too

• Computing power: It is expensive and scarce. If Anthropic cannot secure or allocate enough, it could fall behind or fail to serve customers well.
• Data centers: The filing flags controversy over data centers’ land, water and power use, and the possibility of new rules or charges. That warning landed just before voters across the country backed data center limits in this month’s elections.
• A few big customers: Anthropic relies on a small number of customers and distribution channels, including SpaceX and Microsoft. Losing one could hurt.
• Government conflict: AP reported that the Trump administration labeled Anthropic a supply chain risk in February, and the company sued the Pentagon in March over what it called retaliation for refusing to allow unrestricted military use. Courts have split, and both sides are appealing.

Why a stock filing matters if you never buy the stock

Public companies face disclosure rules that private ones do not. Brian Mulberry, chief market strategist at Zacks Investment Research, told AP that U.S. transparency requirements force IPO issuers to spell out risks clearly so investors can decide whether to take part. That same transparency benefits customers. For the first time, one of the major AI developers has put its own assessment of its risks in a legally binding document, and it will have to keep updating that picture in regular reports once it is public.

The filing also shows the financial pressure behind the AI tools you use. AP noted that Anthropic, OpenAI and SpaceX’s AI unit are all losing money, which has fueled talk of a bubble. Companies that are spending heavily and racing rivals have strong incentives to grow fast, add features and raise prices for premium tiers. Knowing that helps you read product announcements with a clearer eye.

The bigger context

Anthropic has long marketed itself as the safety-focused AI lab. In its filing, the company argues that frontier AI companies must invest in safety and alignment, and presents going public as a step toward reducing AI’s risks. In September, CEO Dario Amodei publicly urged the industry to slow down so safety work could catch up, and warned that swarms of AI agents could become capable of taking over the internet within six to 12 months.

Critics might see a tension between calling for caution and raising record sums to build faster. Supporters might argue that a safety-minded company needs money and scale to have influence. Either way, the filing makes the debate concrete. The co-founders’ voting control and the benefit trust mean public shareholders will have limited power to change the company’s direction, which cuts both ways: it may protect safety commitments from short-term pressure, and it also concentrates control in a few hands.

What to do with this information

A plain note first: this is general information, not investment advice. IPOs can be volatile, and anyone considering buying shares should read the full filing and talk with a licensed financial professional.

For everyday AI users

1. Treat AI answers as drafts. A company warning about its own models’ risks is a good reminder to verify important information, especially health, legal and money questions.
2. Be careful with autonomy. If an app offers to act on your behalf, such as sending emails, booking travel or making purchases, start with low-stakes tasks and review what it did.
3. Read the terms that matter. Look for how your data is used, whether conversations train future models, and how to opt out.

For small businesses building on AI

1. Map your dependencies. Write down which AI vendors your workflows rely on and what happens if one raises prices, changes limits or has an outage.
2. Ask vendors about safety and oversight. Useful questions: Do you publish safety evaluations? Do outside groups test your models? How do you notify customers of major changes?
3. Keep humans in the loop. For anything customer-facing or financial, require a person to approve AI output before it goes out.
4. Watch for hype. A blockbuster IPO will bring a wave of marketing. Judge tools by results in your own business, not by headlines.

What to watch next

The key dates and numbers still to come include the final share price, how much is actually raised, and the valuation the market assigns on the first day of trading. After the debut, Anthropic will file regular public reports, giving outsiders a steady look at its finances and risk disclosures. Watch, too, for how rivals respond, especially OpenAI’s own listing plans, and for whether new state rules on data centers start appearing in AI companies’ risk sections and pricing.

The takeaway

An IPO filing is a sales document and a confession at the same time. Anthropic’s version is unusual because the confession is so long and so blunt about the technology at the center of the business. For anyone who uses AI, the lesson is not panic. It is attention: use these tools with care, keep a person accountable for important decisions, and pay close attention to what the companies building them are now required to say out loud.

Source: The Associated Press. Related AP reports republished by PBS NewsHour: risk factors and the IPO filing.