Sunny Hostin has licensed the right to make synthetic appearances of herself
Sunny Hostin has licensed her image and likeness to AI film and television company Fountain 0 for future digital appearances, according to reporting published August 28. The arrangement may include productions connected with adaptations of Hostin’s novels and represents an unusually visible example of a performer deliberately authorising synthetic use of her identity.
Why visual creators should care
Most AI-likeness stories begin after somebody discovers an unauthorised replica. This starts with a licence. For photographers, filmmakers and agencies, that is a much more useful model: the person’s synthetic identity is treated as a commercial right that needs to be negotiated before generation.
Practical takeaway
Where a performer or client may appear synthetically, record the AI-likeness permission separately from the original photography or filming agreement. Define which productions, appearances, channels and time periods the synthetic licence covers.
What to watch next
Watch what Fountain 0’s actual production and licensing structure reveals about approval rights, compensation, duration and whether Hostin has continuing control over individual synthetic performances.
Sources
OpenAI’s image provenance now has a verification route that production systems can use
OpenAI’s provenance documentation, updated August 27, says supported images generated with ChatGPT, Codex and the OpenAI API carry both C2PA Content Credentials and invisible SynthID watermarks. OpenAI’s verification service can look for either signal, and organisations can integrate supported provenance checks through a verification API. OpenAI stresses that a positive result establishes an OpenAI provenance signal, not that the image is accurate, unedited, legally owned or presented in the correct context.
Why visual creators should care
For publishers and agencies, provenance is more useful when it can be checked automatically during ingestion rather than manually inspecting every image. But the limits matter too: provider provenance can establish origin without establishing copyright ownership or truth.
Practical takeaway
If OpenAI-generated visual material enters a professional pipeline, preserve the original file and record the provenance result separately from rights clearance. 'Verified as generated by OpenAI' does not mean 'cleared for commercial use'.
What to watch next
Watch whether DAM systems, publishers and creative platforms begin using provider verification APIs automatically and whether provenance survives ordinary social-media and client-delivery processing.
Sources
One synthetic-media platform is defining disclosure around whether the audience could mistake the image for reality
Twinen updated its Synthetic Media Disclosure Standard on August 27. The policy requires disclosure when AI-generated or materially altered image, video or other media could reasonably be mistaken for reality, and when material edits change the meaning of a scene by adding or removing people or objects, altering actions or compositing scenes. The rule applies to advertising and sponsored content as well as ordinary posts.
Why visual creators should care
This is a more practical threshold than labelling every use of machine learning. It focuses on whether AI materially changes what viewers believe happened, which is close to the distinction many photographers and agencies need when deciding whether routine AI-assisted editing needs disclosure.
Practical takeaway
Build disclosure around material visual change rather than software names. Removing noise and manufacturing a person who was never photographed are not the same production event.
What to watch next
Watch whether larger social and advertising platforms move towards similar materiality tests and how they treat background replacement, generative object removal and synthetic people.
Sources
A synthetic face can be licensed like a campaign asset instead of treated as an unlimited release
CastNym describes a consent-first marketplace in which verified creators set their terms and brands request a signed scope before generating AI video. Its example licence specifies the campaign, number and duration of videos, channels, geography, time period and fee. Generated output is tied back to the signed agreement through an audit record.
Why visual creators should care
This closely resembles traditional photography and talent usage licensing. The brand does not simply acquire a person’s face forever; it receives permission for a defined commercial use.
Practical takeaway
Synthetic likeness agreements can borrow directly from established usage licensing. Record the campaign, platform, territory, duration, permitted variants and compensation rather than relying on an unlimited AI clause.
What to watch next
Watch whether mainstream model and talent agencies adopt campaign-scoped AI likeness licensing and whether platforms begin embedding licence identifiers into generated video provenance.
Sources
An AI product-photography platform is promising that one brand’s catalogue will not train another brand’s model
Afterlight’s Data Rights Policy, effective August 3, says brands retain ownership of uploaded product photographs, logos and brand assets and own generated imagery. It also states that customer catalogues are not used to train models that other customers can use. The company presents these commitments as contractual guarantees rather than only marketing claims.
Why visual creators should care
Product photographers and agencies routinely handle unreleased products, packaging, campaign assets and brand libraries. A no-cross-customer-training promise can therefore matter as much as copyright ownership in the generated output.
Practical takeaway
For AI product photography, ask whether a no-training promise means no general model training, no other-customer training, or no processing at all. Those are different commitments and should be recorded accurately.
What to watch next
Watch whether larger commercial image platforms adopt similarly plain-language data-rights policies and whether agencies begin making these guarantees a requirement in client procurement.
Sources
Stability AI’s training opt-out shows why ownership and model improvement need separate records
Stability AI’s current Terms of Service say users retain ownership of their inputs and, to the extent legally possible, receive Stability’s rights in generated outputs. Stability may use inputs and outputs to develop and improve its services, including model training, but users can opt out of having their content used for training. Its Acceptable Use Policy separately prohibits infringement of intellectual-property and privacy rights and impersonation without consent or legal authority.
Why visual creators should care
Owning the output does not answer what the provider may do with the inputs. For photographers and agencies uploading reference material, the training setting is therefore a separate part of the rights position.
Practical takeaway
If a client requires a no-training workflow, record when the Stability opt-out was active and which account or service was used. Do not treat output ownership as proof that reference files were excluded from model improvement.
What to watch next
Watch whether Stability moves training controls into more granular project-level settings and whether enterprise agreements provide stronger default protection for client reference material.
Sources
