Tech Oversight Project and Midas Project publish 'The OpenAI Files' governance report

Two advocacy groups, the Tech Oversight Project and the Midas Project, published “The OpenAI Files,” a compiled record of governance concerns about OpenAI drawn from more than 200 sources, including public records and accounts attributed to former employees. The report was timed to OpenAI’s ongoing move away from nonprofit control of its for-profit subsidiary, a restructuring the company was negotiating with its board and investors through 2025.

The document’s central allegations concerned chief executive Sam Altman personally: that he had, at various points, misled board members, engaged in undisclosed conflicts of interest, and resisted investment in safety work relative to product shipping. It also catalogued broader governance complaints — restrictive non-disclosure and non-disparagement agreements that reportedly threatened departing employees’ vested equity, board members with financial ties to the restructuring they were overseeing, and a pattern the authors said showed commitments made when OpenAI was founded as a nonprofit in 2015 being progressively narrowed.

The report’s specific target was OpenAI’s proposed conversion, which the authors argued would remove the profit cap on investor returns, weaken the legal requirement to prioritise public benefit over shareholder interest, and shift ultimate control from a nonprofit board to a conventional for-profit structure. Neither Tech Oversight Project nor the Midas Project has an operational stake in OpenAI; both describe themselves as public-interest watchdogs, and the report functioned as an advocacy document rather than an investigation with independent verification of every claim. It added to a running public argument over OpenAI’s governance that had been building since the board’s brief ouster of Altman in November 2023, and was widely cited in subsequent coverage of the restructuring fight.