Microsoft says AI provenance should survive the advertising workflow
Microsoft Advertising has published dedicated guidance covering AI-generated, AI-manipulated and other synthetic advertising. Advertisers are responsible for obtaining necessary rights, permissions and consent and for providing disclosures where applicable law requires them. Importantly for visual creators, Microsoft also tells advertisers to preserve watermarks, metadata and other provenance information identifying how material was created. Ads may be rejected, restricted or removed for prohibited deepfakes, deceptive synthetic content, unauthorised likeness or voice use, missing required disclosure or interference with required machine-readable provenance.
Why visual creators should care
Provenance is beginning to become a delivery requirement rather than optional metadata. A designer or agency that receives an AI-assisted photograph with Content Credentials can no longer assume stripping that information during production is harmless.
Practical takeaway
Keep a provenance-bearing master and test your advertising export workflow to see whether resizing, transcoding or asset-management systems remove credentials. Do not deliberately clean metadata from AI-assisted creative without checking the destination platform’s rules.
What to watch next
Watch whether Microsoft specifies particular provenance standards such as C2PA and whether rejected advertising begins generating practical guidance about which transformations break machine-readable records.
Sources
New York has seized 12 websites built around turning real people’s photographs into AI deepfakes
The Manhattan District Attorney announced the court-authorised seizure of 12 domain names used to disseminate, publish and sell non-consensual AI-generated intimate imagery. Prosecutors say operators used AI image and video creation tools to transform pre-existing photographs and footage of approximately 1,200 real people into hyper-realistic synthetic sexual material. The victims were overwhelmingly women and included actors, athletes, musicians, politicians, activists and social-media creators.
Why visual creators should care
The enforcement action targets the downstream commercial infrastructure around synthetic likeness abuse rather than simply an individual user making a deepfake. It also demonstrates how an ordinary existing photograph can become the source material for an entirely new category of unauthorised exploitation.
Practical takeaway
For photographers working with recognisable talent, keep releases and source-file records capable of showing what uses were actually authorised. Permission to publish a portrait should not be interpreted as permission for third parties to create synthetic performances or intimate derivatives.
What to watch next
Watch for criminal charges arising from the investigation and whether domain seizure becomes a repeatable enforcement tool against commercial deepfake services.
Sources
UK regulators have banned ads that invited people to turn somebody else’s photograph into sexualised AI video
The UK Advertising Standards Authority has banned five social-media advertisements promoting AI companion, portrait and image-to-video applications. Several promoted the ability to upload photographs of real women and transform them into sexualised synthetic imagery or video. One encouraged users to upload an image of their 'crush'; another demonstrated a clothed woman being transformed into topless video. The ASA said encouraging creation of explicit material from people’s images could cause real-world harm and said future illegal advertisements may be referred to platforms and law enforcement.
Why visual creators should care
The regulatory focus is moving from the final synthetic image to the advertised workflow itself. Promoting a tool by showing how somebody else’s photograph can be sexually transformed without consent is becoming a compliance problem before a user even performs the generation.
Practical takeaway
Agencies marketing generative visual tools should review demonstrations and sample imagery as carefully as the product itself. A promotional example that implies unauthorised likeness transformation can create its own regulatory exposure.
What to watch next
Watch whether the ASA begins coordinating routinely with law enforcement under the UK’s newer restrictions on non-consensual deepfake creation tools.
Sources
An AI rights registry is trying to make consent travel with the digital replica
Footballer Cristian Volpato announced on September 15 that he is using Callandor’s AI Rights Registry to create a verified record governing authorised synthetic uses of his identity. The record covers consent, permissions, licensing terms and approved AI uses involving his likeness, voice, biometric profiles and performance data. An accompanying AI-generated launch video demonstrates an authorised digital twin, and Callandor says the framework is intended to allow approved multilingual brand campaigns while keeping usage restrictions attached to the underlying identity licence.
Why visual creators should care
Traditional releases usually document permission in a contract separate from the media. Digital replicas create a stronger need for permission itself to be discoverable and verifiable when a synthetic asset is generated or licensed.
Practical takeaway
For synthetic talent, build an identity-rights record containing the same practical information you would expect in a serious image licence: approved brands, media, territory, duration, transformations and prohibited contexts.
What to watch next
Watch whether rights registries become interoperable with C2PA or other provenance systems so a generated campaign asset can point directly to the permission governing the likeness inside it.
Sources
Visual-art organisations are drawing different lines between an AI-edited photograph and an AI-generated image
Two current visual-art calls illustrate an increasingly important distinction. The 41st Chelsea International Fine Art Competition opened on September 15 and accepts photography and digital media but prohibits AI-generated art, warning that AI-art submissions may lead to account suspension. The UNFOLD photography exhibition takes a narrower position: fully AI-generated images are prohibited, but photographers may use AI tools as part of their workflow or editing process.
Why visual creators should care
For photographers and visual artists, 'AI prohibited' is no longer a sufficiently precise rule. Generative removal, expansion, retouching and wholly synthetic image creation can receive different treatment even when all involve AI somewhere in production.
Practical takeaway
Before submitting AI-assisted photography to an exhibition, competition or client, record exactly what AI did. 'AI-assisted' may not tell a curator whether the underlying photograph remains eligible.
What to watch next
Watch for organisations to replace broad AI bans with definitions based on source photography, degree of generation and required disclosure.
Sources
AI video generation from professional photographs is moving onto the photographer’s phone
Property-imaging platform Nodalview released an update on September 15 bringing its AI Video workflow to mobile. Property agents and photographers can select existing photographs on location and generate a video directly from the property page rather than returning to a desktop workflow. The accompanying editor now also accepts conventional video sequences such as drone footage, presenter clips and agency graphics alongside generated sequences and includes improved AI captioning.
Why visual creators should care
Generative video is becoming an ordinary extension of a professional photography assignment rather than a separate post-production experiment. That makes the scope of the original photography licence increasingly important: a still photograph may now become source material for motion assets delivered through the same commercial job.
Practical takeaway
For property and commercial photography, specify whether delivered stills may be converted into AI-generated video. If the client controls the workflow after delivery, make that derivative use explicit in the licence rather than leaving it implied.
What to watch next
Watch whether professional photography platforms begin recording which source stills generated each video sequence and whether that relationship appears in exported provenance.
Sources
