A 3D artist is trying again to make stripped attribution and NoAI information part of the AI-training case
Professional 3D artist Austin Beaulier filed an amended complaint against Meta on September 16 after a federal court dismissed his earlier DMCA claims with leave to amend. Beaulier publishes 3D models and photogrammetry scans through repositories including Sketchfab under Creative Commons licences. His case alleges that attribution and licensing information was stripped while files were prepared for Meta's generative-AI training and also points to Sketchfab's machine-readable NoAI tag. The August dismissal found problems with the earlier theory, including whether Meta intentionally removed copyright-management information during file conversion, but allowed Beaulier to replead.
Why visual creators should care
For photographers, illustrators and 3D artists, this gets close to a central provenance question: what legal significance survives when attribution, licence conditions or an AI opt-out are attached to a work but disappear during a technical conversion performed before training?
Practical takeaway
Preserve evidence of the metadata, licence and AI restrictions attached to important work when it is published. A screenshot of a portfolio page is useful, but keeping the original file, licence version and machine-readable restriction provides a much stronger record of what accompanied the work.
What to watch next
Watch Meta's response to the amended complaint and whether Beaulier can plead facts showing intentional removal of copyright-management information rather than an incidental consequence of converting files for training.
Sources
Owning the photograph does not stop a photo-to-3D service from learning from it
3DMyPhoto updated its terms on September 16 for its AI service that converts 2D photographs into 3D models. Users retain ownership of uploaded photographs and resulting 3D models. However, the service receives a worldwide, non-exclusive, royalty-free licence to use, reproduce and modify input material for providing and improving the service, and the terms expressly say this may include anonymously using the material to train AI models. The Shopify-specific terms similarly allow submitted product photographs to be used anonymously to diagnose problems and improve the AI pipeline.
Why visual creators should care
The photographer can retain copyright while the platform simultaneously acquires permission to use the photograph as AI-improvement material. Ownership and training permission are separate contractual questions.
Practical takeaway
Before uploading client product photography to an AI conversion service, check whether your client agreement allows the photograph to be used for model or service improvement. Your own right to process the photograph does not necessarily include the right to contribute it to another company's training pipeline.
What to watch next
Watch whether commercial photo-to-3D services begin offering no-training or enterprise-isolation options for brands that cannot permit product imagery to become improvement data.
Sources
Another AI photo service draws a much narrower line around the uploaded original
FrameLifter's terms, updated September 16, cover an AI service that enhances lighting, exposure and overall quality in property-listing photographs. Photographers must own or have permission to use uploaded images and retain ownership of their originals. FrameLifter says its licence over those photographs is limited to processing them for the enhancement service, while the user owns the resulting enhanced photographs and may use them commercially. The terms do not state that uploaded photographs are licensed for AI training.
Why visual creators should care
Placed beside broader platform licences, the terms show how differently two AI photography services can treat essentially the same act of uploading a professional image. One may receive training or improvement rights; another may limit itself to performing the requested edit.
Practical takeaway
Look for the purpose clause attached to an AI platform's content licence. 'You retain ownership' is not enough. The important words are what the service is subsequently allowed to do with the file.
What to watch next
Watch whether narrow processing-only licences become a selling point for professional photography services handling client-owned or confidential imagery.
Sources
An AI ad platform can now turn the photographs on a product page directly into a finished video campaign
OpenClips announced on September 16 that users can give its system a product URL rather than a prompt, script or shot list. The platform extracts imagery, product features and positioning from the page, creates a storyboard and script and routes individual shots through a network of more than 45 image and video models. OpenClips says paid plans include a commercial licence covering advertising, client work and monetised channels.
Why visual creators should care
The source asset may now enter an AI workflow without anybody manually uploading it. A product photograph already published on a client's website can become input for a new generated video simply because somebody supplies the page URL.
Practical takeaway
When licensing product photography, consider whether permission covers automated extraction and transformation into AI advertising. Website publication rights and permission to create new generated motion assets from those photographs are not necessarily the same thing.
What to watch next
Watch whether automated URL-to-ad systems expose which source photographs were used for each generated scene and which underlying model processed them.
Sources
An AI camera is trying to keep generation inside the boundaries of the photograph that actually happened
Superpose announced its AI-powered iPhone portrait camera on September 15. The system provides live guidance on pose, lighting and camera angle and then generates four alternative versions from a captured photograph. The company says its design deliberately preserves the underlying reality of the moment: it does not move people to invented locations, replace their clothing or substitute the people who were actually present. Instead, generation is focused on choices resembling those a photographer might make around pose and composition.
Why visual creators should care
The workflow sits in the increasingly important space between conventional photography and synthetic image generation. A real photographic event remains the source, but AI generates alternative visual versions of that event.
Practical takeaway
When describing AI-assisted photography to clients or audiences, record what was allowed to change. 'Based on a real photograph' is much more informative when accompanied by boundaries such as pose, framing, lighting, clothing, location and identity.
What to watch next
Watch whether camera and editing applications begin preserving these transformation boundaries as provenance information rather than leaving them only in product descriptions.
Sources
An AI film festival is asking filmmakers for the production record, not just the finished film
The Bochnia International AI Film Festival takes place in Poland on September 17. Its formal rules require qualifying films to have at least 50 percent of their visual or audio content generated or significantly processed with AI and require filmmakers to use at least two different AI tools. Entrants must supply a complete list of those tools and a production diary or making-of record documenting the creative process. Filmmakers retain copyright in submitted work while confirming that they hold the necessary third-party rights.
Why visual creators should care
The festival turns AI disclosure into a production-documentation requirement. Rather than asking creators simply to tick an 'AI used' box, it asks for enough information to reconstruct how AI participated in making the work.
Practical takeaway
AI filmmakers should maintain the tool list and production diary while the project is being made. Reconstructing models, source assets and major generative steps months later is considerably harder than recording them as part of ordinary production.
What to watch next
Watch whether festivals, commissioners and distributors begin standardising similar disclosure packages and whether provenance files such as C2PA records eventually replace part of the manual production diary.
Sources
