Most eyes are on Anthropic's revenue growth, model capabilities, and trillion-dollar valuation expectations as it races toward its IPO, yet few have paid attention to its governance arrangement, which differs sharply from the standard Silicon Valley template. When other tech companies go public, most rely on dual-class AB shares to keep voting rights in the founders' hands, but Anthropic has neither taken that path nor handed power to investors; instead, it has placed the company's actual control in the hands of a group of external trustees who hold no shares and receive no profits—former Federal Reserve Chairman Bernanke is on that list of trustees. This design directly collides with the underlying consensus that capital markets have maintained for years: listed companies should prioritize safeguarding shareholder returns.
Underpinning this arrangement is a two-tier structure combining a Public Benefit Corporation (PBC) with a Long-Term Benefit Trust. As a legal vehicle, the PBC is still a for-profit enterprise that can raise funds, issue shares, list, pay dividends, and compete in the market; its most fundamental change is rewriting the board's legal responsibilities—under its rules, management, when making decisions, need not relentlessly prioritize shareholder returns and may lawfully weigh public safety and social impact. On top of this, Anthropic added a layer of Long-Term Benefit Trust (LTBT) with real power, composed of externally selected experts, who hold no company equity, enjoy no listing appreciation, receive no dividends, and whose compensation only covers their time costs. The trustees hold two core powers: first, the right to replace a majority of board members; second, when the company undertakes major unconventional borrowing, sells more than half of its equity, or transfers core AI models above a set threshold and similar transactions, it must notify the trust five days in advance—many commercial law scholars who have studied these clauses point out that, once implemented, this five-day notification effectively forms a soft veto; if the trust questions whether a transaction poses safety risks, it can block its realization through personnel appointments and removals and sustained pressure.
Zooming out to the global top-tier AI race, governance routes have split into three distinctly different paths. OpenAI's commercial operating entity is likewise a PBC, but at its top level a nonprofit foundation holds control and a for-profit subsidiary beneath it conducts commercial operations; Anthropic's entity remains a for-profit PBC, but it has stripped control out to an independent trust, with the trust holding no equity and only exercising voting rights; Musk-founded xAI was briefly registered as a PBC, then simply abandoned it, reverting to an ordinary commercial company, shedding its mission constraints—all decisions revolve around commercial growth, with safety bottom lines resting entirely on corporate self-discipline. Anthropic sits between the two: it preserves market-based financing channels and will not be tied hand and foot by a nonprofit entity the way OpenAI is, while relying on the independent trust to turn its safety mission into a hard constraint, attempting to balance the freedom of commercial expansion with the public responsibility of AI technology.
Why do large-model companies value corporate responsibility so highly? The root lies in the strong spillover nature of technological risk—under an ordinary commercial company governance framework, capital can easily pressure management to loosen model permissions and launch high-risk businesses for the sake of short-term financial statements. Although the PBC legally gives the board room to place AI safety above profit, it cannot stop shareholders or founders from changing course under enormous commercial temptation, so Anthropic added the LTBT on top, forming a double constraint: the PBC is merely the legal vehicle; what truly holds real power, and is most controversial, is that group of third-party trustees independent of both shareholders and the founding team.
This structure is generating two completely opposing views between the primary and secondary markets. The primary market's bullish logic is clear: as a third-party gatekeeper, the trust can avoid the control struggles common in Silicon Valley and the internal friction between founders and investors, ensuring strategic continuity; amid the global trend of tightening AI regulation, a governance structure with built-in safety constraints is itself a moat of compliance, making it hard for capital to push the team into taking technical risks for quarterly statements, so slow variables like R&D and safety evaluation can continuously receive resources. The secondary market, however, has practical concerns: shareholder rights suffer structural dilution—investors buy shares yet can hardly influence transactions capable of changing the company's destiny, with the switch on major matters held by trustees who are accountable to no shareholders; once the trust vetoes, on safety grounds, a merger or licensing deal that could bring enormous returns, shareholders can only passively accept it. Traditional due diligence can account for revenue, gross margins, and the competitive landscape, but cannot build a valuation model around the trustees' subjective judgments—this gives rise to an unavoidable paradox: the company possesses every trait of a high-growth target, yet the rights carried by its equity are incomplete, and the market will ultimately apply a discount for this deficiency in rights. The primary market bets that long-term upside can accommodate the discount, while secondary market funds, with clear assessment cycles, are naturally more cautious about this governance risk; this cognitive gap will run through pricing battles for a long time after listing.
Viewed over a longer horizon, the significance of this goes beyond the success or failure of Anthropic's own IPO. In the short term, AI market sentiment is hot enough that scarcity will overshadow the governance controversy, and the offering will most likely go smoothly; in the medium term, as share prices fluctuate and major transactions materialize, discussions of excessive trust power and opaque rules will recur, and once trustees' decisions directly clash with investor interests, shareholder lawsuits, institutional repositioning, and valuation reappraisals are all foreseeable plotlines; over the long term, if this mechanism can smoothly run the balance between commercial expansion and risk control, it has the chance to become a new template for top large-model companies entering public capital markets. Capital markets have operated for a century on a foundation built on shareholder-supremacy logic, while AI's risk spillover and social influence are something neither the internet nor the semiconductor race ever had—the old capital rules may not fit this new technology, and Anthropic is using this IPO to run trial and error for the entire industry; where this experiment ultimately leads will define what posture future giant AI companies should adopt as they step into the public market.