专家质疑 Anthropic IPO 的公共利益保障机制
Experts Question Anthropic’s IPO Safeguards for Public Benefit
美联社 11 月 16 日报道,治理专家认为 Anthropic 的 IPO 将检验其使命能否承受盈利压力,其公共利益结构对社会效益如何衡量留有较大裁量空间。Anthropic 11 月 10 日提交的修订版 IPO 注册文件将公共利益目的定义为负责任地开发和维护先进 AI 以造福人类,但长期利益信托不参与日常管理,七位联合创始人持有的 Class F 股份将拥有 50.1% 的一般投票权。Anthropic 已向公共利益基金捐出占全面摊薄股本 5% 的 Class N 股份,联合创始人另承诺捐出个人 Anthropic 股权的 80% 用于慈善。
推荐理由
AI 评分 59
文章拆解了 Anthropic IPO 文件中的信托与创始人控制权安排,读者可据此理解其公益使命在上市后如何被约束。
主题
正文
The Associated Press reported on November 16, 2026, that governance specialists see Anthropic’s planned initial public offering as a test of whether its mission can withstand pressure to deliver profits, with the company’s public-benefit structure leaving wide latitude to decide how social benefit is measured.
Anthropic’s amended IPO registration statement, filed on November 10, 2026, defines its specific public-benefit purpose as the responsible development and maintenance of advanced AI for the long-term benefit of humanity. Delaware law requires its directors to balance that purpose with stockholders’ financial interests and the interests of people materially affected by the company’s conduct.
Sarah Kreps, director of Cornell University’s Tech Policy Institute, said “the real test emerges if and when the question of public benefit becomes costly,” or when it conflicts with making a profit. That distinction matters because the legal structure requires directors to balance competing interests without prescribing which must prevail in a particular business decision.
Trust and Founder Control
The Long-Term Benefit Trust is designed around five financially disinterested voting trustees with expertise in AI safety, national security, public policy and social enterprise. A single Class T share lets the trustees vote as a unit and generally elect and remove the Class T directors who constitute a majority of the authorized board. The share carries no economic rights.
That authority over board seats does not extend to running Anthropic. The filing expressly says neither the trust nor its trustees has decision-making authority over day-to-day management or strategy. Their role is to select directors equipped to balance the company’s three obligations and to support management through advice and guidance.
The founders hold a separate source of control. A single Class F share to be issued to the seven co-founders’ Founder LLC immediately before completion of the offering will carry 50.1% of general voting power, including elections of non-Class T directors, and will have no economic rights. The arrangement divides influence between trustees who can shape most board seats and founders who retain majority voting power on matters submitted to stockholders generally.
UC Berkeley law professor Stavros Gadinis said the trust’s accountability power is consequently limited: trustees who disagree with directors could face a difficult removal process and might be unable to remove a majority. Three directors are elected directly by the trust and three by other shareholders. A seventh cannot be appointed or removed without consent from the board’s nominating committee, which is expected to be controlled by the founders.
Funding Commitments and Measurement
Anthropic contributed Class N shares representing 5% of its fully diluted shares as of June 30, 2026, under the filing’s stated conversion assumption, to its wholly owned Anthropic Public Benefit Fund. The board may issue additional shares representing up to 2% of outstanding shares plus specified convertible and fund-held securities each time the share price rises 50% above the prior milestone. Separately, the co-founders have pledged 80% of their personal Anthropic equity to philanthropic causes.
Anthropic’s October 18, 2026 announcement said it established the fund as an LLC rather than a 501(c)(3) to retain flexibility among forms of giving. None of the fund’s Anthropic equity will be used for lobbying or political giving.
Those commitments define potential resources for public benefit more clearly than they define the result. Kreps said public benefit is likely harder to measure than profit and that competition, including from open-source Chinese models, could pressure AI developers to reconsider guardrails. The measurement question therefore sits alongside the control question: even a well-resourced mission needs a way to judge whether it is being met.
Transparency After the IPO
Gadinis said public-company financial reporting, outside questions and companies’ interest in investor trust could provide an additional accountability mechanism. Required disclosure and scrutiny from investors may generate more information about Anthropic’s decisions than was available while it was private.
The practical test is whether the trust, founders and board use their respective powers when safety and financial interests diverge, and whether public reporting gives outsiders enough evidence to assess those choices. Anthropic’s governance and public-reporting obligations create channels for accountability, but the meaning and measurement of humanity’s long-term benefit remain open to judgment.