OpenAI has implemented a restrictive off-boarding agreement for departing employees, which has caused embarrassment for CEO Sam Altman. The agreement includes provisions that prevent outgoing employees from speaking negatively about OpenAI for the rest of their lives, as well as revoking their vested equity if they criticize the company in a way that diminishes its value. Additionally, employees are not allowed to discuss the terms of their exit, as doing so would also result in the revocation of their equity.
The recent departure of high-profile members of OpenAI’s safety team shed light on these restrictive agreements. Following the release of OpenAI’s flagship product, ChatGPT-4o, co-founder and chief scientist Ilya Sutskever and co-leader of the Superalignment team Jan Leike left the company. Their departures raised speculation that OpenAI may be changing its approach to safety, leading to suggestions that they resigned in protest.
Altman, known for making public statements, has apologized for the missteps in the equity policy. He admitted that the provision regarding the potential cancellation of equity should never have existed in their documents and stated that the company has been working to correct the exit paperwork for the past month. Altman expressed his embarrassment over the situation and encouraged former employees with old documents to contact him for corrections.
However, Altman has faced criticism for contradictions between his words and actions in the past. For instance, he has been accused of promoting AI development in Saudi Arabia, despite advocating for AI safety and signing a letter to the US government calling for caution. It has also been noted that Altman transformed OpenAI from a non-profit into a technocratic startup and accelerated its commercialization, which some view as deceptive.
OpenAI’s history is also relevant to the current situation. Altman was briefly fired by the company’s board last year before being reinstated, and Sutskever, who was on the board at the time, later regretted the decision and joined employees in supporting Altman’s return. Sutskever’s departure from the company is attributed to personal reasons, although it is speculated that his focus on AI safety played a role.
Leike’s resignation was straightforward, with him simply stating, “I resigned.” He later elaborated on his concerns in a series of posts. Altman has stated that the company is working on rectifying the issue with the exit papers. However, former employee Daniel Kokotaijlo surrendered his equity after refusing to sign the agreement, citing a loss of confidence in the company’s responsible behavior regarding artificial general intelligence (AGI).
While Altman has offered to address these matters, the ongoing employee exodus and the questions raised about OpenAI’s changing approach to AI safety make it uncertain how these efforts will unfold.
