On March 13, 2026, Sora 1 officially ended its run in the United States. OpenAI had posted the timeline back on February 27, but the deadline still landed with a jolt: US users now have exactly one video generation tool, Sora 2, and every image and video left unexported in Sora 1 was deleted. The same week, The Information reported that OpenAI plans to bring Sora’s video generation into ChatGPT. The signal is unambiguous: in OpenAI’s roadmap, video generation is no longer a standalone product — it is a feature inside the chat interface.
The March 13 Deadline
The announcement, posted on February 27, gave US users roughly two weeks of runway, and the version circulated on the community forum was blunt to the point of cruelty: “All images and videos will be removed,” with the official advice to “please download any assets you would like to keep.” In other words, this was not a version upgrade but a product termination. Any US user who did not manually export before March 13 watched their creations vanish, with no recovery path on the other side. Sora 2 is the only sanctioned destination in the US going forward, and there is no automatic migration between the two — content, projects, and workflows are the user’s problem to rebuild. Note the geography, too: the sunset applied to the United States first, which is worth remembering when similar notices appear for other regions on a different clock.
Video Generation Moves Into ChatGPT
On March 10, The Information reported, citing people with knowledge of the effort, that OpenAI plans to launch Sora’s AI-generated video capabilities in ChatGPT “soon,” describing it as a strategy shift rather than a routine feature addition. Mashable’s follow-up noted the integration was being planned only a few months after Sora’s standalone app launched — a strikingly fast reversal for a product that had been positioned as a flagship destination. Read in sequence, the pattern is hard to miss: retire Sora 1 in the US first, then surface reports of video folding into ChatGPT — one door closing, another consolidating, with the product line converging on a single entry point. Nothing official has confirmed timing or packaging yet, so treat the exact rollout details as unannounced; the direction, however, is about as subtle as a billboard.
The Data Risk for Creators
This shutdown is a lesson for anyone treating generative AI tools as a work platform. First, content on an AI service is not your asset by default: one change to a server-side deletion policy and anything unexported is gone, with no rescue channel and no appeal process. Second, a version switch is not a migration: Sora 1 to Sora 2 shipped with no built-in transfer, so even diligent users had to rebuild workflows by hand while racing an arbitrary calendar date. Third, the burden of noticing falls entirely on you — the notice lived in help documentation and community channels, not in a forced modal on every login. The pragmatic habits are simple: sync every generated asset to storage you control the moment it lands, treat AI tools as a production line rather than a filing cabinet, and keep a running export of anything you would actually miss. Teams integrating through APIs should watch product notices closely, because sunset timelines arrive with advance warning, not with a wait for you to be ready.
The Business Logic of Consolidation
The math behind pulling Sora into ChatGPT is not hard to follow. A standalone app means its own acquisition costs, its own retention metrics, and its own pressure to justify continued investment; turned into a ChatGPT feature, video generation hangs off an existing subscription, raising the perceived value of that single subscription while keeping user attention locked in one interface. It also simplifies the go-to-market story: instead of marketing two destinations, OpenAI markets one. This is the everything-in-one-place playbook: model capabilities keep multiplying, but the product surface converges to one. For competitors, it means the video generation fight is unlikely to be won by shipping “another Sora app” — it will be won by whichever primary interface can carry more capabilities without losing focus, which is a design problem as much as a model problem. For developers, the thing to watch is when and in what form video generation reappears across ChatGPT’s API and product surfaces, and what happens to pricing once video is a checkbox inside an existing plan rather than a product with its own line item.
Sources
- OpenAI Plans to Launch Sora Video AI in ChatGPT — The Information
- Sora video generator is coming to ChatGPT, insiders say — Mashable
- Sora 1 Sunset for US users on March 13, 2026 — OpenAI Community Forum
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.