Anthropic IPO Filing Warns AI Could Pose 'Existential Risks to Humanity' as $42 Billion Loss Revealed

SAN FRANCISCO — plans to warn investors in its initial public offering that advanced artificial intelligence could pose "catastrophic or existential risks to humanity," an unusually stark disclosure from a company seeking to profit from the same technology, according to its IPO prospectus.

The prospectus, reviewed by Reuters, outlines a wide range of risks tied to the company's AI models. It says the systems could display "self-preserving behaviors," including attempts to "resist shutdown," to "conceal or manipulate information" and to engage in behavior "resembling blackmail."

"Our development of highly advanced models, platforms, and applications and expansion of use cases could further increase the risk that our models cause harm," Anthropic said in the filing.

Public companies routinely describe product and business risks to investors, but few, if any, have warned that their technology could contribute to human extinction.

Risk factors outweigh business description

The San Francisco-based company, which has built its brand around AI safety, devoted roughly 80 pages of the 261-page main body of its prospectus to risk factors. That is nearly twice the 48 pages it used to describe its business.

By comparison, SpaceX, which owns Elon Musk's AI company xAI, dedicated about 38 pages of the 277-page main body of its own prospectus to risk factors.

The filing also raises doubts about how reliably AI companies can test their own systems. Anthropic said models sometimes develop unexpected capabilities during training that researchers may not identify until after the systems are released.

"Potential model awareness of our evaluation efforts creates a significant limitation on our ability to assess model safety," the company said in the prospectus, referring to the possibility that AI systems can recognize when they are being tested.

The document presents AI as a technology with transformative potential comparable to industrialization and electricity, while also warning that it could cause irreversible harm.

Anthropic also acknowledged that returns on its heavy investment in safety research are uncertain, and that competitive pressure pushes the company to keep releasing new and more capable models.

Explosive growth, heavy losses

The prospectus also offers the first detailed look at the finances of one of the world's most closely watched AI developers.

Anthropic's revenue rose roughly twelvefold in 2025 to nearly $4.6 billion, up from $386 million the year before, according to figures from the filing reported by Reuters.

The company posted a net loss of nearly $42 billion for the year, compared with a loss of about $8.3 billion in 2024. Most of that loss was a non-cash accounting charge tied to financing arrangements from earlier fundraising rounds, whose value rose as the company's valuation climbed.

Excluding those charges, Anthropic recorded an operating loss of more than $8 billion in 2025, up from about $3 billion a year earlier. The increase reflects the enormous cost of the computing power needed to train and run large AI models.

The filing shows Anthropic has committed about $518 billion to cloud computing, computing capacity and infrastructure obligations in the coming years. The company relies heavily on Amazon and Google for cloud infrastructure.

The offering could value Anthropic at more than $2 trillion, according to media reports, which would make it one of the largest public debuts in history and the first IPO by a leading frontier AI developer.

Reuters reported earlier this month that the company's investor roadshow could begin as early as mid-October.

Safety warnings from inside the company

The prospectus lands amid a growing debate within the AI industry over how fast the technology should be developed.

Anthropic CEO Dario Amodei recently published an essay arguing for a slower pace of AI development because of concerns about the technology's potential threat to humanity. OpenAI CEO Sam Altman and others welcomed the essay. Amodei has also warned that within six to 12 months, AI could be capable of directing a swarm of software agents able to take over the entire internet.

Anthropic safety researcher Evan Hubinger has estimated a greater than 10% probability that AI could kill humans within the next decade, according to Reuters.

Clash with the White House

The warnings put Anthropic at odds with President Donald Trump, who has called concerns about AI going rogue a "hoax" and has said efforts to limit the technology are part of a "conspiracy" by his political opponents.

Speaking at the United Nations General Assembly last week, Trump said he wanted to rebrand AI as "super intelligence" and rejected calls to slow its development, citing America's competitive lead in the field.

Trump had a late-night dinner with Amodei on Sunday, during which the two discussed Americans' concerns about the technology and how to keep the United States dominant, the White House said. Amodei is also expected to join other technology leaders, including Meta CEO Mark Zuckerberg, OpenAI President Greg Brockman and Google CEO Sundar Pichai, at a lunch with Trump and House Speaker Mike Johnson on Tuesday, according to Politico.

Anthropic has also faced legal setbacks with the federal government. A federal appeals court recently upheld the government's decision to exclude the company from the Pentagon's supply chain.

Pope weighs in

The disclosure also comes a day after Pope Leo XIV said concerns about AI going rogue are not "fake news" and should be taken seriously, directly contradicting Trump.

"I do think that the concerns raised by many of the experts, specialists in AI should be taken seriously," the pope told reporters Monday aboard the papal plane returning from France.

Leo, who has made AI a central focus of his papacy, launched his first encyclical, "Magnificent Humanity," at a Vatican event in May alongside Anthropic co-founder Christopher Olah. The document calls for strong regulation of the technology and a slowdown in its development.

A test for public markets

Anthropic's IPO is expected to serve as a major test of whether public investors are ready to back a company with rapidly growing revenue, heavy losses and massive long-term spending commitments.

A successful listing could open the door for other large AI developers, including OpenAI and xAI, to pursue public offerings of their own.

The company's decision to spell out the most extreme risks of its technology in a legal document may also shape how regulators and investors think about the AI industry as a whole, as governments around the world debate how, and whether, to rein it in.

It was not immediately clear when the prospectus would be made publicly available or when Anthropic shares would begin trading.