Mira Murati leaves OpenAI alongside two research leaders

Mira Murati, OpenAI’s chief technology officer, announced Wednesday that she is leaving the company after more than six years. Her departure coincides with those of Bob McGrew, head of research, and Barret Zoph, vice president of research—three major exits for a company that has put generative models at the center of the tech industry.

Murati was one of OpenAI’s most recognizable executives. Her work covered turning research into products such as ChatGPT, DALL-E and GPT-4, along with the technical decisions involved in deploying them. In May, she was the public face of the launch of GPT-4o, the model with text, image, audio and video capabilities that OpenAI presented as its next step toward making ChatGPT more conversational.

A central figure since 2018

Murati joined OpenAI in 2018, when the organization was still far less known outside technology circles. She became chief technology officer in 2022 and briefly served as interim chief executive during the November 2023 governance crisis, after Sam Altman was initially ousted.

The episode ended with Altman’s return and the renewal of the board, but it exposed tensions over the pace of AI commercialization and the control mechanisms at a company developing increasingly capable models. Murati was one of the few senior leaders to remain in a position of top responsibility throughout the transition.

In the message announcing her decision, she said she wanted time and space to explore new projects. Altman publicly thanked her for her contribution and said the company would arrange a transition for her responsibilities.

Bob McGrew and Barret Zoph are also leaving

The timing amplifies the significance of the news. Bob McGrew led OpenAI’s research organization, a role that connects scientific work with the company’s product strategy. Barret Zoph, the vice president of research, had played a prominent role in post-training models: the stage when a base system is fine-tuned to follow instructions, respond helpfully and reduce problematic behaviors.

They are not the first major departures from OpenAI this year. Ilya Sutskever, the company’s co-founder and chief scientist, announced his exit in May. Jan Leike, who co-led the superalignment team focused on researching how to keep highly advanced AI systems under control, followed shortly afterward. In August, John Schulman, another co-founder, announced that he was joining Anthropic.

Each departure has its own context, and none of the three executives announced today has publicly attributed the decision to a specific conflict. Still, the accumulation matters because OpenAI is competing in a market where research talent is scarce and people who have helped develop these models possess expertise that is difficult to replace overnight.

The challenge of growing without losing technical strength

In less than two years, OpenAI has gone from launching ChatGPT to operating products used by hundreds of millions of people and closing commercial deals with major companies. That expansion requires stronger engineering, infrastructure and sales teams, but it also requires preserving a research leadership capable of deciding which models to train, how to evaluate them and under what conditions to make them available to the public.

The company maintains a dominant position in product popularity, though it faces pressure from rivals including Google, Anthropic, Meta and xAI. These competitors are not just fighting for customers; they are also seeking researchers and executives with first-hand knowledge of the techniques and processes required to build frontier models.

OpenAI’s immediate test will be filling these roles without slowing its roadmap or weakening its safety evaluation processes. Murati, McGrew and Zoph’s departures do not by themselves change ChatGPT’s capabilities, but they do require the company to show that its technical structure can endure beyond a group of leaders who have been decisive to its growth.