Anthropic's IPO Filing Warns Its Own Models Could 'Cause Harm' as Losses Mount

Anthropic is preparing to go public, and the filing that will take it there includes an unusual admission: the company says its own development plans could increase the chance its models cause harm. The Verge reported that a preview of the IPO paperwork pairs that risk language with rising losses and proposals that would let current leadership keep control after shares start trading. Anthropic is reportedly eyeing a valuation near $2 trillion, a figure that would put it among the most valuable companies ever to list.

The safety warning is not boilerplate. According to the preview, the filing states that the company's plans for building more capable systems could "further increase the risk that our models cause harm." That is a striking thing for a company to tell prospective shareholders, because it frames the core product as a potential source of catastrophic damage rather than a neutral tool. Anthropic has built its public identity around that tension, positioning itself as the lab most willing to talk about what could go wrong.

The financial picture is less flattering. The filing reportedly lays out mounting losses, which means Anthropic would arrive on public markets as a revenue-generating but heavily unprofitable business. That combination is familiar from the current AI buildout: enormous spending on chips, data centers and talent, with the payoff still projected rather than proven. A $2 trillion target would require public investors to accept that gap on faith.

Governance is the third piece. Leadership has proposed structures that would retain power after the listing, a setup that echoes the dual-class arrangements used by founder-led technology firms for years. For Anthropic, which has restructured its corporate form before to accommodate its safety commitments and its relationship with backers including Amazon and Google, the ask raises a direct question: what exactly are public shareholders buying a vote on?

The broader significance is what this filing does to the sector's story. AI labs have raised private capital at extraordinary valuations while promising that safety and scale can advance together. Anthropic's own risk disclosure, if it reads as reported, undercuts the clean version of that pitch and gives regulators and skeptical investors a company-authored document to quote back at the industry.

What to watch next is the full filing and its timing. The numbers that matter are the size of the losses, the terms of the share structure, and whether underwriters price the offering anywhere near the reported $2 trillion mark. If they do, Anthropic will have convinced public markets to fund a business that says, in writing, that its success could make things worse.