A multi-model image platform is making creators check two contracts instead of one
NanoGPT’s revised Terms of Service take effect for existing users on September 14. The service provides access to multiple AI providers, including image and video models. NanoGPT says users retain their existing rights in inputs and, as between the user and NanoGPT, own generated output. Images, video and audio may be used commercially. NanoGPT also says it does not train on users’ inputs or outputs. But those promises do not override the separate terms of whichever underlying model actually handles a request. NanoGPT explicitly tells users to verify provider-specific restrictions covering commercial use, retention, training and other practices.
Why visual creators should care
For professional visual creators, one interface can now hide several different contractual environments. Two images generated from the same account but with different models may not carry identical usage rights or data-handling rules.
Practical takeaway
Record the underlying image or video model with important client generations, not only the aggregator you logged into. Before commercial delivery, verify the provider terms that applied to that model and retain them with the project record.
What to watch next
Watch whether multi-model platforms begin freezing provider terms at generation time or attaching model-specific rights information to individual assets so creators can reconstruct the contract later.
Sources
C2PA is moving from individual newsroom experiments toward a standard implementation framework
A technical paper presented at IBC on September 13 by Media Cluster Norway, CBC/Radio-Canada and IPTC sets out a standardised framework for implementing C2PA media provenance across news organisations with different technical resources and workflows. The same IBC session paired that work with Reuters’ approach to open newsroom infrastructure, where C2PA sits alongside standards for storing, exchanging and processing professional media.
Why visual creators should care
Content Credentials become more useful to photographers and filmmakers when publishers implement them consistently. If every newsroom signs, edits and verifies media differently, provenance becomes another fragile proprietary workflow rather than a portable professional record.
Practical takeaway
Creators supplying signed photographs or footage should ask whether the receiving organisation preserves parent provenance when adding captions, metadata, crops and other edits. The useful goal is a chain that grows rather than a credential that gets replaced.
What to watch next
Watch for publication of implementation guidance from the participating organisations and whether the framework feeds into IPTC’s forthcoming media-provenance best practices.
Sources
A photo-to-video platform is screening the photograph before it lets the generation happen
Picmovi’s new AIGC Use Policy took effect September 12 across its image-to-video, video-generation and AI-image tools. Commercial creative use is permitted, but users may not make unauthorised use of a real person’s face, likeness or voice or use copyrighted works, trademarks or logos without the required rights or a valid legal exception. Picmovi says both prompts and uploaded images are screened on its servers before generation, with blocked or reported jobs potentially retained in limited form for enforcement.
Why visual creators should care
Safety and rights controls are increasingly being applied to the source photograph rather than only to the prompt or final synthetic file. That makes uploading the reference itself a meaningful legal step in the workflow.
Practical takeaway
Do not upload a client portrait, model photograph or branded asset into a photo-to-video service simply because you are allowed to possess the file. Confirm that your licence or release covers the intended synthetic transformation.
What to watch next
Watch whether visual services begin distinguishing ordinary photography releases from explicit permission for face animation, synthetic performance and image-to-video generation.
Sources
An artist platform is turning 'we will not train on your work' into a contractual covenant
NewBohemia’s current legal terms contain an Artist Work Protection Covenant stating that the platform does not sell artists’ submitted work or verification materials, does not license those materials to third parties for their own AI, machine-learning, dataset or model-development use, and does not itself use them as training or model-development data. The covenant also says these protections cannot simply be removed retroactively from previously submitted covered material through an ordinary later terms update.
Why visual creators should care
Most creator-platform promises about AI training exist as changeable policies. A contractual covenant that also limits later retroactive alteration attempts to make the no-training commitment part of the durable bargain under which an artist uploaded the work.
Practical takeaway
When choosing portfolio or verification platforms, compare contractual AI-training language rather than relying only on marketing claims such as 'artist friendly'. Save the version of the terms that applied when important work was uploaded.
What to watch next
Watch whether larger portfolio and marketplace services adopt similarly durable no-training clauses and whether creators begin treating these protections as a competitive feature when deciding where to host work.
Sources
A synthetic-advertising tool is designing real-person impersonation out of the workflow
IntactShot’s September policy describes a commercial system that turns a product photograph into a short video featuring a synthetic presenter. Users cannot upload a face, perform a face swap or clone a voice. The presenter library consists of generated people that the company says were not modelled on real individuals, and it keeps a provenance record for each presenter. Product photographs must be owned by the user or used with authorisation. The platform also requires publishers to comply with applicable synthetic-media disclosure rules.
Why visual creators should care
Most AI safety policies say a user must not misuse a real person’s likeness. IntactShot illustrates a different approach: removing the capability to supply that likeness at all. For commercial agencies, product photography can then remain the rights-bearing source asset while the synthetic human is deliberately separated from a real model.
Practical takeaway
When choosing synthetic-presenter tools for commercial work, check whether they permit user-supplied faces and voices. A closed, documented synthetic cast can remove an entire layer of talent-clearance risk, although the product photograph and advertising claims still need their own rights checks.
What to watch next
Watch whether synthetic talent platforms begin providing downloadable provenance and licence records for each avatar and voice so agencies can attach those records to campaign files.
Sources
Verifiable creator identity is being added to the provenance chain, not just the file history
A second provenance project presented at IBC on September 13 combines C2PA with Creator Assertions Working Group identity standards in an open-source publishing workflow. Dawn Technology and Media Campus NL are testing the system with Dutch broadcaster VPRO. The project signs media for provenance while also linking it to digitally verifiable identity, and the implementation is being released as open source to lower adoption barriers for other publishers.
Why visual creators should care
Knowing that a file was signed is only part of the trust problem. Professional photographers and filmmakers also need a reliable way to say who stands behind that signature. Connecting creator identity with the provenance record can make attribution substantially more useful than an anonymous technical credential.
Practical takeaway
Where provenance matters commercially, think about identity and attribution as part of the credential design. A signed file that cannot reliably connect back to its creator or publisher leaves an important part of the evidence chain unresolved.
What to watch next
Watch the VPRO pilot for evidence of how CAWG identity survives editorial changes and whether similar creator-identity assertions become available in mainstream photography and video software.
Sources
