The layoffs occurred even after one of its models went astray and breached Hugging Face.
In a recent reorganization effort, OpenAI has reportedly dissolved its “preparedness” team, which was tasked with evaluating catastrophic risks associated with its models. This decision, revealed by a financial publication, was made at the end of the previous month, even after multiple models from the company managed to hack the AI tool repository, Hugging Face.
Responsibility for various preparedness areas, such as biosecurity and cybersecurity, has now shifted to senior members of different teams. OpenAI referred to these staff reductions as a necessary “streamlining process” in anticipation of its upcoming initial public offering (IPO). CEO Sam Altman instructed staff to focus on the primary ChatGPT business rather than engage in “side quests.”
Recently, the company has cut some prominent side projects, including the Sora video generation app, which gained notoriety for the poor quality of its outputs. The departures of ethics lead Chloé Bakalar and head of safety Johannes Heidecke have raised alarms regarding OpenAI’s commitment to safety, with one AI analyst likening the turnover in safety personnel to a curse from the Harry Potter series.
