Anthropic's S-1 filing devotes nearly a third of its length to risk factors, warning advanced AI could pose "catastrophic or existential risks to humanity," including models capable of resisting shutdown, concealing information, and behaviour resembling blackmail. The financials underneath are just as striking: revenue rose twelvefold to $4.6 billion in 2025, but $13 billion in operating expenses produced an operating loss of over $8 billion. Anthropic has committed roughly $518 billion to future infrastructure, and is targeting a listing valuation above $2 trillion on the Nasdaq this autumn, ahead of SpaceX's $1.78 trillion mark. Nearly a quarter of 2025 revenue came from just two clients.
The disclosure itself is the interesting governance signal, not just the risk it names. Choosing to spell out existential risk this candidly, rather than the regulatory minimum, is a real choice, and whether it reads as genuine caution or a liability shield ahead of a landmark listing will likely depend on who's doing the reading.