OpenAI Policy Expert Departs, Citing Industry Accountability Gaps
Gretchen Kreuger, a policy researcher at OpenAI, announced her resignation this week, using a post on X to flag what she described as insufficient attention to decision-making processes and accountability within the AI industry. Her departure adds to a string of exits from the company's safety and policy ranks.
In her online thread, Kreuger said the choice to leave was not easy, but she emphasized the need for improvements in areas like transparency, documentation, policy enforcement, and the careful use of technology. She also pointed to the importance of addressing impacts on inequality, rights, and the environment, stating that these concerns affect how the future is shaped and by whom.
Kreuger's exit comes at a time when OpenAI has faced scrutiny over several issues, including allegations that it used actress Scarlett Johansson's voice without permission for a product, and reports from Vox about the company's practice of requiring exiting employees to sign strict non-disclosure and non-disparagement agreements under threat of losing vested equity.
She is not alone in leaving. Ilya Sutskever, the company's former chief scientist, and Jan Leike, a researcher who worked on the now-disbanded Superalignment team, have also departed. That team was responsible for ensuring that future advanced AI systems would not act against human interests, and its dissolution has raised questions about OpenAI's commitment to safety.
Concerns Over Internal Accountability
Kreuger's remarks also touched on how tech companies can undermine those who seek accountability, by dividing people who raise concerns. She said she cares deeply about preventing such division, a statement that resonates with the recent departures of safety-focused staff.
The pattern of exits among employees involved in safety and policy work suggests potential friction between those roles and the company's broader direction. While OpenAI has not publicly commented on Kreuger's departure, the cumulative loss of key figures in safety-related positions may signal ongoing internal tensions.
Observers note that the decisions made by leading AI companies have wide-reaching implications, yet the processes for making those decisions remain largely internal. Kreuger's warning highlights a growing concern about the lack of external oversight and the need for more robust governance mechanisms in the industry.
