Legal scholars say the AI group’s disclosures on existential risk, not investor demand, are now the swing factor in what bankers hope will be the largest listing in history
Anthropic confidentially filed a draft S-1 with the Securities and Exchange Commission on June 1, and people familiar with the process told Reuters on September 4 that the public prospectus is now expected in late September, with marketing beginning in mid-October and the listing completing days before the November midterm elections. Bankers and investors have floated a valuation near $2 trillion, which would make it the largest initial public offering on record; as part of the process, Anthropic is finalizing a $15 billion revolving credit facility.
The timeline has been complicated less by market appetite than by what the company must say about its own product. John Coffee, a securities-law professor at Columbia Law School, told The Economist that Anthropic may need to amend its registration statement to limit its exposure to securities-fraud claims, after a former Anthropic alignment researcher, Jacob Coxon, resigned publicly on September 8 warning that frontier AI could pose an extinction-level risk within the decade. Evan Hubinger, Anthropic’s alignment-science lead, backed the substance of the warning days later, putting his own estimate of catastrophic risk above 10% within ten years and saying the company does “not yet have a plan to solve alignment for superintelligence.” Coffee’s reading cuts against the instinct that more candor raises liability. He said Anthropic’s existing safety-first governance structure may mean the risk is already disclosed enough in its S-1 paperwork to blunt fraud claims if something goes wrong later. Plaintiffs’ lawyers argue the exposure runs the other way: investors who bought into a “$2 trillion” AI story could later claim they were misled by what the company chose to leave out.
Anthropic’s response so far has been to lean into disclosure rather than away from it. Chief executive Dario Amodei published a roughly 3,800-word essay on his personal site on September 12 arguing the industry should deliberately slow the rate at which it improves model capabilities, citing recursive self-improvement, where models increasingly help build their successors, as a specific new risk. He proposed embedded third-party evaluators, from organizations such as METR, to verify AI labs are keeping pacing and safety commitments. OpenAI’s Sam Altman endorsed the idea publicly. The essay landed a day before the Financial Times reported, on September 13, that Anthropic had told investors its adjusted operating income would be positive for a second straight quarter, with gross margins above 80% before revenue shared with distribution partners including Amazon and before the cost of training new models.
The underlying growth is what makes the liability question matter to a listing this size. Anthropic’s annualized revenue run rate reached $65 billion at the end of July, up from roughly $9 billion at the end of 2025, after quarterly revenue rose fourteenfold year on year to $11.5 billion, Bloomberg reported on August 17. OpenAI has chosen the opposite disclosure strategy. Founder and chief executive Sam Altman told Fortune on September 12 that going public now would be “an ill-advised moment,” pushing that company’s own listing into 2027 or later.
Separately, a Moody’s Ratings report published September 14 estimated the broader AI buildout will require $110 billion of new US power generation, roughly 45 gigawatts, through 2030, a capital need that will fall on data-center operators and utility ratepayers well beyond any single company’s IPO. For Anthropic’s bankers, that is a financing problem for later. The more immediate one sits on page one of the prospectus, where the company still has to decide how much of what its own safety team believes belongs in print before the roadshow begins.