Interview, Fireside Chat
OpenAI just tried to kill off its nonprofit owner – and failed
- The Safety and Security Committee (SSC) holds broad discretion to constrain or halt AI releases, potentially without limit if volunteers define safety and security autonomously, creating a scenario where the four volunteer directors may struggle without dedicated supporting staff.
- The viability of the current governance structure relies heavily on the performance of specific board members; hiring management, adding a second non-overlapping director with deep AI safety expertise, or appointing a dedicated SSC leader are viewed as positive signals, while reliance on volunteers or conflicted appointments are negative indicators.
- External oversight may intensify, with the expectation that Attorneys General (AGs) will treat current communication requirements as a floor rather than a ceiling and may mandate public announcements on mission compliance or progress.
- Shareholder mechanisms exist to hold the Public Benefit Corporation (PBC) accountable, allowing a group holding at least 2% of outstanding shares to sue the board for breaching fiduciary duties if the OpenAI charter or safety mandates are violated.
- Financial projections suggest the Public Benefit Corporation (PBC) valuation could increase 10x over the next 15 years if the nonprofit succeeds significantly, though the actual value of the nonprofit's warrant remains uncertain, ranging from a meaningful approximation of profit caps to a meaningless "fig leaf."
- Future operational models include possibilities for the nonprofit to step down to become a non-overlapping director, sell down its stake to fund philanthropy, or commandeer PBC resources via support agreements, though no dedicated revenue share was finalized in the initial proposal.
- Governance timelines and requirements may evolve to include specific reporting frequencies (annual, biennial, or ad-hoc), mandatory 21-day advance notice of changes, and scheduled meetings with the nonprofit board or senior staff, potentially increasing transparency obligations.
- Microsoft's Intellectual Property rights are anticipated to remain with the PBC until 2032, allowing for the possibility of independent commercialization of AGI by Microsoft or by the PBC itself, while the PBC management could focus on the mission rather than exclusively on shareholder value.
- Risks are elevated if AGI is developed rapidly, as the nonprofit's ability to intervene for public good might be lost, and if the SSC proves toothless, investors may proceed despite safety concerns or the board may face litigation for insufficient oversight after a failure occurs.
- The arrangement is characterized as potentially better than the December baseline or competitors like Google and XAI, yet it is viewed as a bad deal compared to an idealistic baseline where OpenAI prioritizes humanity's benefit over being first or securing a power grab.