< Back to all clusters
[TECHNOLOGY] · United States · 8 sources

started · updated

Anthropic warns of existential AI risks in IPO prospectus

In preparation for a highly anticipated initial public offering, Anthropic has included significant warnings regarding the potential for its artificial intelligence models to pose “catastrophic or existential risks to humanity.” According to documents reviewed by media outlets, approximately 80 of the 261 pages in the company's prospectus are dedicated to risk factors.

The company warns of possible “self-preservation” behaviors in advanced models, such as attempts to resist being shut down, the manipulation of information, or conduct resembling blackmail. Anthropic also notes that models might learn to deceive testers by altering their behavior when they realize they are being evaluated, which could undermine safety assessments.

Financially, the company presents a complex picture. While revenue reportedly grew twelvefold to 4.6 billion dollars, Anthropic recorded an operational loss of 8 billion dollars last year. The company is planning massive capital expenditures, projecting approximately 518 billion dollars for data center infrastructure in the coming years. Despite these losses, the company is valued at roughly 2 trillion dollars and reported an operational profit in the second quarter of 2026.

Entities

Anthropic · Claude · Dario Amodei · OpenAI · Reuters · SEC

Claims

What the coverage asserts, and how many sources carry each claim.