Sora reaches its shutdown date, which makes archiving part of the AI-video workflow
OpenAI’s published discontinuation date for the Sora API is September 24. The Sora web and app experiences had already closed on April 26. OpenAI recommends exporting Sora content and says that after discontinuation and the end of any final export period it is able to offer, data associated with Sora use will be permanently deleted.
Why visual creators should care
AI filmmakers can possess contractual rights in a generated clip while still depending on somebody else’s infrastructure to retain the actual asset. A model or service disappearing can therefore become an asset-management problem rather than simply a software problem.
Practical takeaway
Do not treat an AI platform library as your production archive. Download masters, source images, reference footage, prompts and important generation records into storage you control.
What to watch next
Watch how production teams migrate old Sora workflows and whether replacement video platforms provide stronger project-export and long-term archive tools.
Sources
Google is making provenance part of every generated Vids clip
Google announced on September 23 that Gemini Omni video generation is now available in Google Vids to anyone with a Google Account. The system generates 1080p video, supports controls including scene duration and transitions, and can extend generated scenes. Google says every AI-generated clip created through the feature includes a digital watermark.
Why visual creators should care
Generated video is moving from specialist AI tools into everyday production software. Making provenance marking automatic rather than optional reduces the chance that disclosure disappears simply because somebody forgot to switch it on.
Practical takeaway
When AI video is used in a professional project, record both the visible final edit and whatever machine-readable or invisible provenance signal the generator attaches. Check whether that signal survives subsequent editing and export.
What to watch next
Watch whether Google exposes more information about the watermark to downstream platforms and whether edited or recompressed Vids footage remains identifiable after leaving Google’s workflow.
Sources
An AI image marketplace is making creators responsible for checking the model licence behind the picture
Pixerr’s September 23 terms cover a marketplace selling photographs, AI-generated images, digital downloads and physical products. Creators uploading AI imagery must confirm that they comply with the terms of the AI tool used and have the right to commercialise the resulting content. Pixerr separately warns that AI-generated material may not qualify for copyright protection and makes no guarantee that it does. Its content rules also prohibit creators from presenting AI-generated work as entirely human-made.
Why visual creators should care
A marketplace can allow an AI image to be sold without establishing that the seller owns copyright in it. Permission to commercialise an output, copyright ownership and disclosure are three separate questions.
Practical takeaway
Keep the model and its applicable commercial-use terms with the asset record. Do not tell a buyer that an AI image carries exclusive copyright merely because the marketplace lets you sell it.
What to watch next
Watch whether AI marketplaces begin showing model, licence and generation provenance directly on product listings rather than leaving those records entirely with the seller.
Sources
An AI image library is licensing what it can without pretending that clears everything inside the picture
BudgetPixel’s September 23 terms place its public AI-generated image library under Creative Commons Attribution 4.0, but carefully qualify that licence as applying only to the extent BudgetPixel actually holds rights in the material. It warns users that generated imagery may resemble another party’s artwork, trademark, person or place and does not warrant that those third-party rights are cleared. Free anonymous generations may also be published with their prompts into the public library under the same licence.
Why visual creators should care
A licence attached to an AI image does not magically clear every right represented inside that image. Copyright in the file, likeness rights, trademarks and underlying protected visual material can sit on different legal layers.
Practical takeaway
When using supposedly free or Creative Commons AI imagery commercially, check what the licence actually covers. A platform cannot grant you rights it never possessed.
What to watch next
Watch whether more AI stock libraries start qualifying their licences explicitly around third-party likeness, trademark and source-material risks.
Sources
AI product photography can be commercially usable without being exclusive
FoodShot AI updated its terms on September 23 for its generated food and product imagery service. Users keep their rights in uploaded photographs and give the service a limited licence to process them. Compliant generated outputs may be used for menus, delivery listings, websites, advertising, social media and print without attribution. FoodShot does not claim ownership of those outputs, but warns that other users may receive similar imagery, that copyright protection is not guaranteed and that users remain responsible for whether the image fairly represents the food or product being sold.
Why visual creators should care
Commercial permission is not the same as exclusivity. An agency can be permitted to use a generated product photograph while having little or no ability to prevent another user receiving something similar.
Practical takeaway
For campaign-critical imagery, ask whether you need permission to use the image or actual exclusivity. If exclusivity matters, a generic AI-output licence may not provide what the client thinks it is buying.
What to watch next
Watch advertising and marketplace rules around generated product imagery, particularly where an AI image makes food, clothing or another physical product look materially different from what the customer receives.
Sources
Photography competitions are increasingly asking for evidence of capture rather than trusting the JPEG
The current reFocus World Photo Annual is open to professional and non-professional photographers across 22 categories. AI-generated imagery is not admissible. Published competition information also says organisers may request the associated RAW file or other evidence establishing ownership and non-AI photographic origin.
Why visual creators should care
As generated imagery becomes harder to identify visually, photography competitions cannot reliably establish origin by looking at the finished JPEG. The evidential burden is beginning to move backwards through the workflow to the camera original.
Practical takeaway
If competitions, editorial publication or documentary credibility matter to your work, retain RAW files and an understandable edit trail. The camera original is becoming useful evidence rather than merely an editing resource.
What to watch next
Watch whether major photography competitions standardise RAW verification, C2PA capture records or defined limits for generative editing rather than relying on broad no-AI wording.
Sources
