Anthropic has formally warned investors that its technology may pose "existential risks to humanity," according to a long-awaited initial public offering prospectus the company circulated with a small group of partners in recent days.

The AI startup, valued at nearly $1 trillion and led by chief executive Dario Amodei, devoted almost a third of its lengthy S-1 filing to detailing "risk factors." Among them: the potential of increasingly advanced AI models to manipulate, blackmail and exhibit other unpredictable behaviors.

The company's decision to put extinction-level risk language at the center of its IPO pitch is unusual, even by the standards of an industry that has grown fond of catastrophic warnings. Most S-1 filings treat risk factors as legal boilerplate, buried deep in the document and drafted primarily to shield issuers from shareholder litigation. Anthropic has inverted that convention, elevating the most dramatic scenarios to headline status in a document whose core job is to sell shares.

The filing also outlined more prosaic commercial risks. The most concrete: an extreme concentration of Anthropic's customer base, with close to a quarter of revenue last year coming from just two clients, according to people familiar with the filing.

That concentration figure may matter more to prospective investors than the existential language. Two customers generating roughly 25 percent of revenue is a structural vulnerability for any company, and it speaks to a broader pattern across the frontier-AI sector, where a handful of large enterprises and cloud providers account for the bulk of spending on advanced models. If either of Anthropic's top two clients cut usage, renegotiated pricing, or shifted workloads to a competitor, the revenue impact would be immediate and material.

The customer-concentration risk also reflects the economics of frontier AI development. Training and serving large models requires massive compute budgets, and the pool of organizations willing to pay for access to top-tier models at scale remains small. Anthropic competes for that limited pool against OpenAI, Google and a growing set of open-weight alternatives, which puts pricing pressure on exactly the customers that dominate its revenue base.

The prospectus circulation marks a concrete step in what has been one of the most anticipated IPO processes in the technology sector. Anthropic, founded in 2021 by former OpenAI researchers, has raised billions from strategic investors and has been the subject of persistent speculation about when it would tap public markets. The document was shared with a small group of partners in recent days, according to the report, suggesting the company is moving through the mechanics of a public offering rather than merely floating the idea.

For investors, the filing presents an unusual trade-off. The existential-risk disclosures align with Anthropic's long-stated positioning as a safety-focused AI lab, and they may serve a dual function: reinforcing the company's brand among customers and regulators who care about responsible AI, while providing legal cover if the technology's behavior becomes a liability. But the same language also frames the product itself as potentially unmanageable at scale — a framing few companies in any industry have volunteered about their own core technology.

The more conventional risk factors cut in the opposite direction. Customer concentration is measurable, near-term and specific to Anthropic's current book of business, unlike the speculative harms described elsewhere in the document. Investors pricing the offering will have to weigh a documented revenue dependency on two clients against a valuation approaching $1 trillion, a multiple that assumes rapid diversification of the customer base in the coming years.

Anthropic has not disclosed a timeline for a public offering, a target valuation for the listing, or the identity of the two clients behind its revenue concentration. Whether the company tempers the existential framing, or the concentration figures, in a wider public filing will signal how it intends to balance its safety-first identity with the growth narrative public-market investors will demand.