Anthropic establishes the Long-Term Benefit Trust
A special stock class gives independent trustees, holding no equity, the right to elect a majority of Anthropic's board within four years.
- Labs & people
- Notable
Anthropic disclosed the structure of its Long-Term Benefit Trust, a governance mechanism the company said it had been developing since its founding to keep commercial pressure from overriding its stated mission. The Trust holds a special class of stock, Class T shares, carrying no equity or profit share, which gives its trustees the power to elect and remove a growing proportion of Anthropic’s board of directors — reaching a majority within four years of the Trust’s formation, on a schedule tied to funding and time-based milestones.
The five initial trustees were selected for expertise Anthropic judged relevant to its stated concerns about advanced AI — national security, public policy, AI safety research and social welfare — and were required to hold no Anthropic equity and take no share of its profits, being compensated only for their time. Anthropic framed the arrangement as a complement to its status as a Delaware public benefit corporation, which already permitted its directors to weigh a public-benefit mission alongside shareholder returns; the Trust added an outside check with actual board-composition power, rather than relying solely on directors’ discretion.
The design was explicitly a response to the possibility that transformative AI could produce catastrophic harms, and to the argument — sharpened a year later by OpenAI’s board crisis — that ordinary venture-backed governance gives investors and executives every incentive to under-weight low-probability, high-severity risks. Critics, including some observers on LessWrong and elsewhere, questioned how much practical power the Trust would have in the near term, given its board influence was designed to phase in gradually and could in principle be diluted or amended by the company’s other stakeholders under some circumstances.
The Trust subsequently added trustees including Reed Hastings in 2025, and later Richard Fontaine, Mariano-Florentino Cuéllar and Ben Bernanke, giving the arrangement a public roster of appointments that became one of the more closely watched aspects of Anthropic’s corporate structure relative to OpenAI’s and other labs’ governance arrangements.