SAN FRANCISCO: Anthropic is preparing for what could become the largest initial public offering (IPO) in history. However, its own prospectus warns that advanced artificial intelligence could pose serious risks to humanity.
The San Francisco-based AI company is reportedly seeking a valuation of $2 trillion. Its roughly 300-page IPO filing presents an ambitious vision for AI while outlining major financial, technological and governance risks.
Reuters reviewed the confidential prospectus this week.
Nearly one-third of the document discusses risk factors. That is more than twice the space devoted to explaining Anthropic’s business.
The filing highlights several contradictions. Anthropic says AI could transform the global economy, but it also warns that the technology could threaten human existence.
The company argues that safer AI requires greater computing power and more capable systems. Yet it acknowledges that increased capability could make those systems harder to control.
AI Safety Depends on Greater Power
Anthropic’s central challenge is how to make increasingly powerful AI systems safe.
The company says its technology could improve quality of life and transform industries worldwide. However, it also warns that advanced systems could pose catastrophic or existential risks if people do not manage them properly.
Margaret O’Mara, a University of Washington history professor who studies the technology economy, said AI companies present their products as both transformative and dangerous.
She noted that this differs from the promises made by companies during the dotcom boom.
Anthropic’s prospectus describes a future of “Powerful AI” that could outperform top human experts across nearly all cognitive fields.
Such systems could work with other AI models, operate laboratory equipment and control robots. They could also run for weeks with little or no human supervision.
The filing also points to possible advances in personalised financial advice, drug discovery and scientific research.
However, Anthropic warns that advanced models may sometimes act against their developers’ intentions. They could also enter other companies’ systems or display behaviours such as resisting shutdown, concealing information or manipulating people.
The company says increasingly capable systems could be misused, become misaligned with human values or escape effective control.
It warns that such systems could damage institutional decision-making, cause widespread harm and weaken public trust.
Heavy Losses and Massive Spending Commitments
Anthropic’s ambitions come with substantial financial costs.
The company lost more than $50 billion over the two years ending in 2025. It also has more than $500 billion in spending commitments for the coming years.
By comparison, SpaceX reported a $4.2 billion loss over the same period, according to the Reuters report.
Anthropic investor Byron Deeter of Bessemer Venture Partners sought to ease concerns about the confidential filing.
Speaking to CNBC on Tuesday, Deeter said people were reacting to a draft. He also praised Anthropic for its disclosures.
Anthropic did not respond to Reuters’ request for comment.
Founders Could Retain Significant Control
The prospectus also raises questions about Anthropic’s governance structure.
The company says concentrating leadership and capital could help it manage AI risks. At the same time, it warns that concentrated power can threaten society.
Anthropic is asking public investors to trust its seven founders with overseeing the company. That structure could limit shareholders’ influence and ability to monitor leadership.
The filing describes the founders as particularly suited to protect the company’s mission. However, it acknowledges that their decisions could conflict with financial interests or business performance over the short, medium or long term.
Chief Executive Dario Amodei addressed the challenge during an appearance at the United Nations last week.
He said no leader, company or country could manage the global opportunities and threats posed by AI alone.
Big Tech Partners Bring Both Growth and Risk
Anthropic’s growth also depends heavily on major technology companies.
Amazon, Google, Broadcom and Microsoft together accounted for 47% of Anthropic’s 2025 revenue through sales and partnerships, according to the filing.
The company also relies on major technology firms for computing capacity.
However, Anthropic warns that these partners could compete with it directly or restrict access to essential services.
The company has also committed to pay for certain computing arrangements even if it abandons a lease or does not use the services.
It says any reduction, repricing or termination of third-party computing capacity could harm its business.
AI Could Expand Wealth but Concentrate Power
Anthropic’s prospectus presents AI as a potential source of global prosperity. It says the technology could help developing countries reach living standards enjoyed by wealthier nations.
But the company also warns that widespread AI deployment could concentrate wealth and influence.
Such concentration, it says, could harm society and destabilise the geopolitical order.
Meanwhile, concerns raised by AI researchers have prompted politicians from both major US parties to consider slowing the technology’s development.
US President Donald Trump has dismissed potential AI harms as a “hoax.”
Amodei and leaders from Google DeepMind, OpenAI and xAI have also called for a slowdown in AI development to focus on safety.
Their position contrasts with the traditional technology-sector approach of expanding rapidly to capture markets.
