Reuters is putting AI editing around verified footage rather than replacing the source with generated video
Reuters announced a partnership with CuttingRoom on September 12 that connects Reuters video with CuttingRoom’s browser-based ShortCut editing system through Reuters’ Model Context Protocol server. Editors can use plain-language instructions to locate material, make cuts, mix audio, correct colour, add captions and create versions for different platforms. Reuters says the system allows verified Reuters material and a newsroom’s own footage to coexist in the same workflow, while individual newsrooms retain control over the AI interaction, their data and the editorial rules applied to the system.
Why visual creators should care
This is a different form of professional AI adoption from simply generating replacement footage. The trusted source video remains central while AI operates on selection, editing and packaging around it. That makes the editorial record of what the system changed particularly important.
Practical takeaway
For AI-assisted video editing, preserve the verified source file and record the AI operations separately. The fact that the underlying footage is authentic does not automatically establish that every later caption, cut, colour adjustment or contextual presentation is editorially accurate.
What to watch next
Watch whether Reuters and CuttingRoom expose operation-level provenance or C2PA records showing which transformations were automated and which were approved by an editor.
Sources
A permanent injunction in India treats an AI-generated face and voice as part of the person’s protected identity
The Delhi High Court has permanently restrained defendants from misusing journalist Rajat Sharma’s name, image, voice, likeness and other personality attributes, including through AI-generated deepfake videos and manipulated imagery. The final decree was made on August 24 and received renewed public reporting on September 12. Sharma and India TV had challenged Facebook pages, YouTube channels and other accounts that allegedly used AI-manipulated versions of his identity alongside unauthorised use of India TV intellectual property. The order provides a mechanism for additional infringing URLs to be reported to Meta and Google for removal.
Why visual creators should care
The protection is not limited to copying an existing photograph. The order treats fabrication of new audiovisual material using a recognisable person’s face or voice as capable of infringing personality and publicity rights.
Practical takeaway
For commercial synthetic-person work, obtain permission for the identity use itself rather than assuming ownership of a photograph gives permission to manufacture new statements, endorsements or performances featuring the person.
What to watch next
Watch whether Indian courts extend similar relief more systematically beyond celebrities and public figures, particularly where ordinary people have no established commercial publicity value.
Sources
Regulators are starting to ask why platforms allowed advertisements for AI tools built around manipulating real photographs
San Francisco City Attorney David Chiu has sent Meta a cease-and-desist demand following reporting and research into advertisements promoting AI nudification services on Facebook, Instagram and Threads. The material included AI-generated abusive imagery and, according to the reporting, some advertisements used or manipulated photographs of real minors. Meta says it prohibits this material and removed offending advertisements, while the city is demanding stronger prevention and information about how the advertisements passed review.
Why visual creators should care
The enforcement question is moving beyond the developer of the image generator. Advertising and social platforms can become part of the rights chain when they distribute services whose commercial proposition depends on making non-consensual synthetic imagery from photographs of real people.
Practical takeaway
Platforms and agencies should treat advertisements for likeness-manipulation tools as a higher-risk category. Reviewing only the ad copy is insufficient where the underlying product is designed to alter photographs of identifiable people.
What to watch next
Watch whether San Francisco escalates the demand into formal enforcement and whether Meta changes its advertiser screening for nudification and other synthetic-likeness services.
Sources
Provenance tools are beginning to reconstruct a media asset’s history even when the original chain has broken
Fraunhofer IDMT is demonstrating an expanded Content Verification Toolbox at IBC 2026 in Amsterdam. Alongside forensic and provenance technologies, the system includes automated provenance tracking designed to identify reused content, connect related versions and reconstruct the genealogy of audiovisual media across distribution channels. Fraunhofer describes this as extending trust beyond the original C2PA-signed publication by examining how content has subsequently travelled and changed.
Why visual creators should care
Professional photographs and video rarely remain in one pristine file. They are clipped, transcoded, screenshotted, reposted and stripped of metadata. If provenance systems work only while the original credential stays perfectly intact, much of the real distribution chain becomes invisible.
Practical takeaway
Preserve C2PA where possible, but do not make an intact credential your only verification method. Keep originals, hashes, publication records and authoritative metadata that can help reconnect derivatives to the source if the embedded provenance disappears.
What to watch next
Watch independent evaluation of genealogy reconstruction after common transformations such as screenshots, social-media recompression, cropping, subtitles and generative edits.
Sources
Creators are discovering AI versions of themselves advertising products they never agreed to endorse
New reporting published September 12 documents influencers and online creators discovering deepfake advertisements built from their photographs and video presence. The synthetic advertisements place recognisable creators into commercial endorsements for brands or products with which they have no relationship. The problem is particularly acute for creators because their face, style and perceived authenticity are themselves part of the commercial asset they sell to legitimate advertisers.
Why visual creators should care
For photographers, models and creators, likeness rights increasingly have a direct commercial dimension. A synthetic endorsement does not merely copy a photograph; it can consume the credibility and audience relationship associated with the person depicted.
Practical takeaway
Where talent is photographed for brands, specify whether AI replicas, synthetic performances and derivative endorsements are permitted. Traditional usage language covering an advertising photograph may be too vague for a campaign capable of manufacturing entirely new appearances by the model.
What to watch next
Watch lawsuits and advertising-platform policies for clearer distinctions between ordinary image reuse, manipulated advertising and entirely synthetic performances created from a person’s online visual footprint.
Sources
Adobe Stock makes the generator’s rights position part of the contributor’s responsibility
Adobe’s current Stock guidance for generative-AI submissions, updated September 2, requires contributors to hold all rights necessary to license generated images, vectors and videos commercially and to check that the AI tool they used permits that downstream licensing. Generated submissions must be labelled under Adobe’s AI-content rules. Adobe also prohibits contributors from using the names of artists, real people, fictional characters, copyrighted works and government agencies in generative-AI prompts, titles and keywords submitted to Stock.
Why visual creators should care
A generator allowing commercial use does not automatically mean an image is suitable for sublicensing through a stock marketplace. The contributor remains the person making the legal representation that the asset can be sold to customers.
Practical takeaway
Keep the applicable generator licence with every commercially significant AI-stock submission. Before uploading, review prompts and references for people, brands, characters and protected creative works rather than checking only the finished pixels.
What to watch next
Watch whether Adobe begins requiring model-level provenance or generator identification alongside its existing AI-content label, particularly for video and heavily reference-led generations.
Sources
