The OpenAI Files is an investigative compendium published on 2025-06-18 by The Midas Project and The Tech Oversight Project. It compiles documented concerns about governance practices, leadership integrity, and organizational culture at OpenAI, and describes itself as "the most comprehensive collection to date" of such material. The publishers state editorial independence, noting they took no funding from xAI/Musk, Anthropic, Meta, Google, or Microsoft.
Summary of findings
The report is organized around four findings.
The first concerns OpenAI's restructuring. The report states that OpenAI is eliminating its investor return caps, previously set at 100x, and transitioning from a capped-profit to a for-profit structure, which the publishers characterize as a departure from the organization's original nonprofit mission.
The second concerns CEO integrity. The report compiles concerns about Sam Altman's leadership voiced by multiple senior leaders, including allegations of psychologically abusive behavior, and assembles testimony from figures involved in the November 2023 board ouster and its aftermath.
The third concerns transparency and safety. The report argues that "Safety culture and processes have taken a backseat to shiny products," and documents the pattern of pressure on departing employees through non-disparagement agreements (NDAs), a practice OpenAI later partially rescinded after reporting on it.
The fourth concerns conflicts of interest. The report describes investor pressure as a driver of structural changes counter to OpenAI's founding principles and documents relationships between Altman's personal investments and OpenAI decisions.
Reform proposals
The report sets out a reform vision built on three elements: responsible governance, including board reform and restoration of board independence; ethical leadership, including accountability for founders; and shared distribution of benefits, reverting toward the nonprofit-mission principles.
Reception and use
The report has served as source material for the Senator Elizabeth Warren letter to OpenAI (January 2026) and comparable congressional inquiries. It functions as a companion to the Sutskever deposition in Musk v. Altman and to Helen Toner's testimony, and was among the most-cited critical references on OpenAI in 2025–2026.
Relationships
- supports: Critique lines in (Source: washingtonpost.com) (overlapping governance concerns), AI Whistleblowing.
- contradicts: The Intelligence Age (Altman), Abundant Intelligence (Altman) (implicit tension with Altman's public vision).
- depends-on: OpenAI.
- related: Sam Altman, Ilya Sutskever, Helen Toner, Miles Brundage.
Provenance
- Primary: The OpenAI Files