- OpenAI’s head of science initiatives — a role held by the company’s former product chief — and the leader of its Sora video generation team have both left the company, Bloomberg reported on April 17, 2026.
- The departures add to a sustained pattern of senior leadership exits at OpenAI that accelerated in 2024 with the departures of CTO Mira Murati, co-founder John Schulman, and chief scientist Ilya Sutskever.
- Sora, OpenAI’s text-to-video diffusion transformer model, has been available to ChatGPT Plus and Pro subscribers since December 2024 and competes directly with Google DeepMind’s Veo 2 and Runway’s Gen-3 Alpha.
- No successors for either role have been announced, and OpenAI had not issued a public statement as of April 18, 2026.
What Happened
Two senior OpenAI executives left the company this week: the head of science initiatives — a position previously occupied by the company’s former product chief following an internal reorganization — and the team lead for Sora, OpenAI’s text-to-video generation model, according to Bloomberg. The exits were reported on April 17, 2026, and represent the latest in a series of leadership transitions as OpenAI works to consolidate an organizational structure that has grown substantially since 2023. OpenAI did not comment publicly on either departure by the time of publication.
Why It Matters
The departures extend a documented pattern of senior attrition at OpenAI. In 2024 alone, the company lost CTO Mira Murati, who left in September; co-founder John Schulman, who departed in August to join Anthropic; and chief scientist Ilya Sutskever, who left in May to co-found Safe Superintelligence Inc. The loss of product and science leadership at this rate raises practical questions about continuity on actively developed products, particularly those — like Sora — that have not yet reached a stable feature set.
Technical Details
Sora uses a diffusion transformer architecture — referred to internally by OpenAI as a spatiotemporal transformer — that jointly models spatial and temporal information across video frames. The model generates video clips of up to 20 seconds at up to 1080p resolution and became available to ChatGPT Pro subscribers at higher generation volumes than Plus tier users when it launched in December 2024. OpenAI has publicly acknowledged limitations including difficulty maintaining consistency in complex physics simulations and multi-entity motion choreography. The departure of the Sora team lead introduces uncertainty around the timeline for the model’s next capability release, which has not been publicly scheduled.
Who’s Affected
Developers who have integrated Sora via OpenAI’s API into production video workflows are most directly exposed to any slowdown in product iteration or support. The science initiatives function, now without a named leader, typically coordinates longer-horizon research that feeds future product cycles — a gap that could affect OpenAI’s competitive position against Google DeepMind, whose Veo 2 model has been available to enterprise customers since early 2025, and Runway, which has continued releasing incremental updates to its Gen-3 Alpha model.
What’s Next
Bloomberg’s report did not name successors for either departing executive, and OpenAI’s next anticipated public event — where product roadmap updates are typically disclosed — has not been formally scheduled. The company’s restructuring of its product portfolio is ongoing, and the shape of the science initiatives function under new leadership, once announced, will indicate how OpenAI intends to balance applied product work against longer-term research priorities.
