Microsoft secretly launched a test of OpenAI’s unreleased GPT-4 model in India without required safety approvals, a breach that intensified board concerns about CEO Sam Altman’s leadership months before his dramatic firing, according to a Wall Street Journal investigation.
The unauthorized test marked the first real-world deployment of the revolutionary AI code and occurred without approval from the joint safety board established between Microsoft and OpenAI to evaluate risks before product releases. Most critically, OpenAI’s independent board members were kept completely in the dark about this violation.
According to the WSJ report, the breach only came to light when an OpenAI employee stopped a board member in a hallway after a marathon six-hour meeting. Neither Altman nor board president Greg Brockman had mentioned the incident during the meeting, despite having multiple opportunities to do so.
This episode became part of a larger pattern of alleged deception that ultimately contributed to Altman’s firing in November 2023. Board members Helen Toner, Tasha McCauley, Adam D’Angelo, and chief scientist Ilya Sutskever had grown increasingly distrustful of Altman’s management style and transparency.
“Things like ChatGPT and GPT-4 were meaningful shifts toward the board realizing that the stakes are getting higher here,” Toner told the Journal. “The board needs to be functioning well.”
The GPT-4 safety breach wasn’t an isolated incident. During a separate meeting, Altman had claimed three controversial GPT-4 enhancements had received safety board approval, but when pressed by Toner, only one had actually been cleared, the WSJ reported.
Board members also learned belatedly that Altman personally owned OpenAI’s Startup Fund, which had been publicly described as “managed” by OpenAI, further eroding trust in his leadership.
These revelations help explain why the board cited Altman’s failure to be “consistently candid” when announcing his termination. However, the board’s inability to communicate these specific concerns publicly contributed to the swift employee backlash that ultimately forced Altman’s reinstatement just days after his firing.
