Anthropic is pursuing what could be a record-setting IPO, aiming for a valuation around US$2 trillion, underscoring a central paradox: safer AI requires more powerful models, yet greater model power elevates risks to humanity.

CEO Dario Amodei emphasizes governance challenges in public appearances, arguing that global collaboration is essential to harness AI opportunities while mitigating risks.

The filing notes AI’s potential to raise living standards in developing regions, but warns that improper power consolidation could destabilize society and geopolitics.

Anthropic counts on major tech partners—Amazon, Google, Broadcom, and Microsoft—for revenue and compute power, representing about 47% of 2025 revenue, while cautioning these partners could constrain or compete with the company.

A core governance paradox arises: Anthropic advocates concentrating leadership and capital among its seven founders to reduce AI risk, yet such concentration could weaken investor oversight and public accountability.

The IPO prospectus cautions about self-preserving and manipulative behaviors in advanced AI and the risk that misalignment could erode institutional trust and cause widespread harm.

Nearly one-third of Anthropic’s 300-page S-1 filing is devoted to risk factors, detailing existential and systemic AI risks that dwarf the sections on the business itself.

The company frames AI as capable of dramatically boosting the economy and quality of life, while also posing potential catastrophic or existential harms if misaligned or uncontrolled, highlighting the safety-vs-power tension.