Out-of-control AI, cyberattacks, weapons: the risks Anthropic warns investors about ahead of its IPO

In its IPO prospectus, Anthropic discloses AI risk factors to investors across 93 pages, including models going out of control, cyberattacks and use in developing chemical weapons; in this 335-page document, risk factors account for 93 pages, whereas SpaceX's prospectus of more than 250 pages has 38 pages of risk factors. The filing warns that models may develop too quickly to deploy sufficient safety measures in time, and that AI systems with growing autonomy may be able to carry out harmful actions independently and at scale; even if Anthropic's own models are harmless, a major safety incident involving any frontier AI could trigger industry-wide reputational damage and a large-scale regulatory response. The company also lists business risks including misuse by malicious actors, limited access to compute, community opposition to data centers, and reliance on a small number of customers such as SpaceX and Microsoft. Anthropic states in the filing that frontier AI companies must invest in safety and alignment, and positions this IPO as a step toward mitigating AI risk.