Skip to content

Goldblatt, The Inverted Benefit Corporation: OpenAI and the Limits of Mission Governance

Andrew Muchin Goldblatt (UCLA) has posted The Inverted Benefit Corporation: OpenAI and the Limits of Mission Governance. Here is the abstract:

In October 2025, OpenAI converted from a nonprofit-controlled LLC to a nonprofit-controlled Delaware public benefit corporation, culminating in a two-year governance crisis and producing the most elaborate mission-control design in the frontier AI sector. This paper argues that the resulting structure generates real friction but fires on the wrong triggers. 

Delaware’s PBC statute was built for for-profit entities adding a public-benefit purpose. OpenAI deployed the form in the opposite direction: a mission-driven entity adopting the PBC to legitimize profit. That directional inversion causes three statutory mechanisms to misfire. The protective function, designed to shield mission-minded deviation from a profit baseline, now shields profit-ward deviation from a mission baseline. Section 102(b)(7) exculpation immunizes the very decisions a mission-controlled entity most needs to police. And the interaction between a diffuse public-goods mission and a newly legitimate profit motive leaves no benchmark against which directorial balancing can be tested. 

Even a tightened PBC regime would not reach the risks that matter most: sustained commercial pressure that never triggers a formal checkpoint, sovereign compulsion that exceeds any private actor’s capacity to resist, and unilateral executive action that bypasses governance entirely. The February 2026 confrontation between Anthropic and the Department of Defense supplies the stress test. OpenAI, with the most elaborate mission-governance overlay in the sector, did not visibly resist the government’s demands; Anthropic, with a thinner structural overlay but a leader willing to absorb existential commercial risk, did. What distinguished the two outcomes was leadership conviction, not corporate form. 

Structure cannot prevent governance failure, but it can make failure visible, costly, and subject to external response. That visibility creates the predicate for the federal regulatory frameworks that do not yet exist.