Back in late 2023, OpenAI’s board of directors shocked everyone by firing CEO Sam Altman out of blue. Now, a report by The New Yorker sheds light on what was happening inside the company in the months leading up to that decision. During a meeting in December 2022, Altman told board members that key features in the upcoming GPT-4 model had already been approved by a safety panel. This was meant to reassure the board that proper checks were in place before launch.
However, Helen Toner, a board member and AI policy expert, decided to verify this. She asked for documentation and what she found told a different story. Some of the most controversial features, including the ability to fine-tune the model for specific uses and to deploy it as a personal assistant, had not actually received safety approval.
Also read: OpenAI CEO Sam Altman tried to scam US govt for billions, here is how
Also, as board member Tasha McCauley was leaving the meeting, an employee quietly asked if she knew about the breach in India. Microsoft had already released an early version of ChatGPT there without completing a required safety review. During many hours of briefing, Altman had not informed the board about this breach. ‘It just was kind of completely ignored,’ Jacob Hilton, an OpenAI researcher at the time, said.
Inside the company, some researchers believed priorities were shifting in the wrong direction. Researcher Carroll Wainwright described it as a ‘continual slide toward emphasising products over safety.’ Former executive Jan Leike wrote to the board, warning, ‘OpenAI has been going off the rails on its mission.’ He added, ‘We are prioritising the product and revenue above all else, followed by AI capabilities, research and scaling, with alignment and safety coming third.’
Soon after this, the board decided to fire Altman.