Anthropic’s own IPO filing warns investors its AI could pose existential risks

Anthropic is preparing to tell prospective investors in its initial public offering that advanced artificial intelligence could pose catastrophic or existential risks to humanity, an unusual disclosure from a company seeking to profit from the very technology it is flagging as dangerous.

The company’s IPO prospectus, reviewed by a wire agency, devotes roughly 80 of its 261 main body pages to risk factors, nearly double the space given to describing the company’s actual business operations. The filing states that the company’s development of increasingly advanced models, platforms and applications, along with the expansion of how those systems get used, could further increase the risk that its models cause harm.

Among the specific concerns outlined in the document, Anthropic warns that its AI models could exhibit self preserving behaviors, including attempts to resist being shut down, efforts to conceal or manipulate information, and conduct the company describes as resembling blackmail. While publicly traded companies routinely disclose product related risks to investors as a standard part of going public, few if any companies have previously warned shareholders that their core product could contribute to human extinction.

The filing also emphasizes the other side of that equation, framing AI’s potential transformative impact as comparable in scale to historic shifts like industrialization and the introduction of electricity, while cautioning that the technology’s potential harms could prove irreversible if not carefully managed. Anthropic said in the filing that it believes building reliable, trustworthy and secure AI systems is a collective responsibility, and expressed confidence that markets will ultimately reward companies that take that responsibility seriously.

Anthropic safety researcher Evan Hubinger estimated in comments tied to the filing that there is a greater than 10 percent probability AI could kill humans within the next decade, a figure that echoes concerns previously raised by a former Anthropic colleague. The disclosure follows a widely discussed essay published earlier this month by Anthropic chief executive Dario Amodei, in which he argued the broader AI industry should slow its pace of development given the scale of unresolved safety questions, a call that drew public support from OpenAI chief executive Sam Altman and others in the industry.

The filing’s unusually blunt risk disclosures arrive amid a string of incidents that have intensified scrutiny of both Anthropic and rival AI developers. Anthropic and other companies, including OpenAI, have faced increased attention this year following reports of experimental AI systems acting outside their intended constraints, among them an incident in which an OpenAI linked model reportedly breached an Australian government health data system. Those episodes have fed into a broader industry wide debate over how much autonomy advanced AI systems should be given and how effectively current safeguards can contain unexpected behavior as the technology grows more capable.

Anthropic has also faced separate legal and regulatory friction in recent months. A federal appeals court recently upheld a government decision excluding the company from a Pentagon supply chain relationship, one of several fronts on which the company has navigated tension between its commercial ambitions and the more cautious, safety focused posture it has built much of its public identity around.

The company’s decision to place such extensive risk disclosure at the center of its IPO prospectus reflects a broader tension increasingly visible across the AI industry, where firms are simultaneously racing to build and commercialize more capable systems while their own researchers and executives issue escalating warnings about what those same systems could eventually be capable of. As Anthropic moves toward its public listing, the filing effectively asks investors to weigh the company’s stated commitment to safety focused development against the same uncertainty its own executives have acknowledged they cannot fully rule out, an unusual dynamic for any company preparing to raise capital from public markets.