Anthropic Files for $2 Trillion IPO With Existential AI Risk Warning as Tech Giants Converge on White House

Anthropic has filed an IPO prospectus targeting a $2 trillion valuation, which would make it the largest public offering in history, while explicitly warning potential investors that its own technology could pose catastrophic or existential risks to humanity. The prospectus also reveals a $518 billion AI infrastructure buildout tied largely to non-cancelable contracts and documents sharply rising costs alongside wider losses, even as the company has grown substantially over the past year. SEC Chair Paul Atkins called the existential risk disclosure "an unusual type of disclosure," and at least one analyst suggested Anthropic was using the doomsday language strategically to deter open-source rivals.

The filing arrived on the same day OpenAI's annual recurring revenue was reported to be approaching $70 billion, sharpening attention across the industry on AI valuations and the competitive stakes between frontier labs. OpenAI separately delayed a new AI model after internal safety testing found it became too persistent in completing tasks and failed to meet the company's safety bar, according to the Wall Street Journal.

Anthropomorphic CEO Dario Amodei was among a group of top tech executives, including Nvidia's Jensen Huang, Amazon's Jeff Bezos, Microsoft's Satya Nadella, Meta's Mark Zuckerberg, AMD's Lisa Su, and Elon Musk, who attended a White House AI luncheon with President Trump and House Speaker Mike Johnson on Tuesday. Trump said he would sign a new executive order on AI, expressed a desire to maintain the U.S. AI lead, and told reporters he does not want to disrupt the technology too much. The White House simultaneously launched America.gov, an AI-powered government portal using Google's Gemini and Grok that consolidates information from roughly 29,000 federal websites and allows users to ask questions in plain language.

On the safety and hardware front, Nvidia unveiled its Open Agent Safety Platform, a containment architecture for autonomous AI agents that strips permissions by default and restores access only when needed. Jensen Huang, speaking with CNBC, argued that AI containment is an engineering problem and that companies should simply not release systems they cannot control. More than 100 organizations are working with the platform, including Microsoft, Anthropic, IBM, Hugging Face, and SpaceXAI, with OpenAI notably absent. AMD also captured attention with an $8.2 billion deal to acquire Fei-Fei Li's World Labs AI startup, drawing praise from Wall Street analysts as AMD pushes further into physical AI.