YouTube adds automatic AI labels in Studio

• YouTube moved its AI disclosure label out of the description and into the main viewing surface, directly below long-form players and as an overlay on Shorts, according to YouTube's update summary and the official blog post.
• Starting in May, YouTube said it will auto-label videos when its systems detect significant photorealistic AI use that a creator did not disclose, per the YouTube update and YouTube's post.
• Creators can still change an incorrect AI disclosure in Studio, but labels stay permanent for content made with Veo or Dream Screen and for media carrying fully generative C2PA metadata, according to the YouTube policy summary.
• The company also said the label alone does not affect recommendations or monetization, as YouTube's announcement spells out.
• OpenAI's Sora shutdown started as a vague goodbye in the Sora shutdown notice, then later solidified into an April 26 app cutoff and a September 24 API cutoff in OpenAI's help article.

You can read YouTube's full post, skim the HN thread where people immediately worried about false positives, and compare that with OpenAI's later Sora FAQ, which quietly turned a fuzzy shutdown announcement into a real export deadline.

The biggest product change is simple: the AI label is no longer buried in the description for photorealistic or meaningfully AI-altered work. YouTube's post says long-form videos now show the label directly below the player, while Shorts get the label as an on-video overlay.

YouTube also collapsed these cases into one visible format. Unrealistic, animated, or only slightly altered content still keeps its disclosure in the expanded description, which creates a sharper split between obvious creative stylization and content YouTube thinks needs an at-a-glance warning.

The second change is YouTube's use of what it calls new internal signals to detect undisclosed significant photorealistic AI use. If the system flags a video, YouTube says it can apply the label automatically even when the creator left the disclosure blank in upload flow, per the YouTube policy update.

That landed exactly where you would expect. In the HN discussion summary, one commenter predicted the detector would get things badly wrong, while another argued the platform is already so flooded with AI video that some automated pass is necessary.

The official override is narrow but important. YouTube's blog says creators can update an incorrect disclosure status in Studio, but not in every case.

YouTube named two exceptions where the label stays locked:

• content made with YouTube's own AI tools, including Veo and Dream Screen
• content carrying C2PA metadata that marks it as fully generative AI

That detail matters more than the headline. The new system is not just a detector plus appeal button. It is also a metadata policy, and YouTube is treating some provenance signals as strong enough to make the disclosure permanent.

The company also stated that the label by itself does not change recommendations or monetization eligibility, according to the policy post. That leaves the immediate effect squarely in viewer context, not distribution penalties, at least on paper.

OpenAI's first Sora message was almost all sentiment and almost no operations. The shutdown post thanked creators, said more details would come later, and promised future guidance on app and API timelines plus preserving user work, as the Sora discussion summary notes.

That uncertainty is what made the HN thread feel less like a normal product sunset and more like a retreat from AI video as a standalone destination. In the HN core summary, the discussion framed Sora's closure as a pullback from the idea that video generation needed its own creator app and social layer.

The early thread also pulled in outside reporting. One comment cited a Wall Street Journal report saying OpenAI was discontinuing a developer-facing Sora offering and would not support video features inside ChatGPT either, a claim surfaced in the Sora discussion highlights.

The later documentation finally filled in the missing dates. In OpenAI's Sora discontinuation FAQ, the company says the Sora web and app experience ended on April 26, 2026, while the Sora API is scheduled to end on September 24, 2026.

The same help page says users can export their data and that OpenAI will email users if a final export window is offered before access fully closes. That is more concrete than the original goodbye post, which left both preservation and cutoff timing unresolved.

Put next to YouTube's labeling update, the contrast is stark. One company is pushing AI video disclosures into the player surface, while another has moved its experimental video app into shutdown-and-export mode.