Photoshop is making the generative edit part of the layer history
Adobe’s August 27 Photoshop update introduces an optional AI Assisted Editor with a natural-language prompt bar, Prompt to Edit, Generative Fill, Generative Expand, Generative Upscale, object removal and AI Markup. Adobe also brings Prompt to Edit into the standard Pro Editor. Crucially for production records, Adobe says every Prompt to Edit result becomes a generative layer in the layer stack rather than replacing the original image destructively.
Why visual creators should care
For photographers and designers, AI editing is moving from a separate specialist feature into the ordinary Photoshop workflow. A layered file can therefore become useful evidence showing which parts of the image were generated, what remained untouched and where human editing continued afterwards.
Practical takeaway
For important client or editorial work, retain the layered Photoshop file rather than only the flattened export. Generative layers can help document where AI entered the workflow and distinguish generated changes from conventional adjustment layers.
What to watch next
Watch whether Adobe exposes model, prompt and provenance information more clearly at layer level and whether exported Content Credentials preserve enough of that history for a client to inspect later.
Sources
Reuters treats AI detection as one clue, not proof that an image is real or fake
Reuters published a detailed account on August 28 of how its global visual-verification team authenticates photographs and video sourced from the public. Journalists identify and contact the original creator, examine metadata where available, compare imagery with weather records, satellite and street-view imagery, look for corroborating eyewitness material and then scan files using several AI-detection tools. Reuters explicitly says those detector tools are not foolproof and may fail to produce a firm conclusion.
Why visual creators should care
The workflow is a useful counterweight to the growing temptation to treat an AI-detector percentage as a verdict. Professional visual evidence still depends heavily on origin, context and corroboration.
Practical takeaway
If the authenticity of an image matters, work outward from its source. Keep the original file, identify who made it and retain contextual evidence. Use AI detection as an additional check, not the only one.
What to watch next
Watch whether newsroom verification increasingly incorporates C2PA or camera-origin credentials alongside the existing source, metadata and corroboration workflow.
Sources
The EU says ordinary assistive AI editing is not the same thing as generating synthetic content
The European Commission’s current Article 50 guidance says generative AI providers must add machine-readable marks to generated or manipulated synthetic content. It also clarifies that the marking obligation does not apply where an AI system performs an assistive function for standard editing. The rules apply from August 2, although some systems placed on the market before that date have until December 2 for the provider-side marking requirement.
Why visual creators should care
For photographers, the distinction is important. Using AI-powered editing software does not automatically mean every processed image falls into the same regulatory category as a synthetic or materially manipulated photograph.
Practical takeaway
Record what the AI actually did rather than simply noting that an AI-powered application was used. Noise reduction, ordinary enhancement and synthetic reconstruction can create very different disclosure consequences.
What to watch next
Watch how regulators and platforms interpret 'standard editing' when tools such as object removal, face repair and generative reconstruction begin to overlap with ordinary retouching.
Sources
Stocksy’s rule shows why an acceptable Photoshop edit can still make a photograph ineligible for a library
Stocksy’s current contributor guidance says it does not accept AI-generated photography, illustration, video or animation. The restriction also covers generative AI used as an editing tool. Stocksy specifically lists features such as Photoshop Generative Fill and says contributors may not use generative systems even for object or logo removal.
Why visual creators should care
Professional workflows are governed by more than software terms. Adobe may allow a photographer to make a generative edit, but a stock library can still decide that the resulting file no longer meets its contributor rules.
Practical takeaway
Check the delivery destination before using generative retouching. If the same photograph may be delivered to clients and submitted to stock, keep a conventionally edited version separate from any generative variant.
What to watch next
Watch whether stock libraries introduce more granular distinctions between synthetic generation, generative retouching and assistive machine-learning tools such as denoising or masking.
Sources
A visual AI platform can let you generate an image without guaranteeing you can safely commercialise it
Proxima Pictures’ current AI Policy says generated or edited outputs may be synthetic, inaccurate, non-unique, misleading or legally restricted. It does not guarantee that output is copyrightable, non-infringing or commercially suitable. The policy also prohibits non-consensual sexual deepfakes, deceptive impersonation and other abusive synthetic-media uses.
Why visual creators should care
Commercial permission to use a tool and legal certainty in the output are not the same thing. The platform can provide generation capability while leaving the creator responsible for copyright, likeness and other third-party rights in the finished visual.
Practical takeaway
Do not turn 'commercial use allowed' into 'commercially cleared'. For important output, review recognisable people, brands, protected characters and other potentially protected material separately.
What to watch next
Watch whether smaller visual generators begin adding optional clearance or indemnity layers similar to those emerging at major stock platforms.
Sources
Scalable makes provenance preservation mandatory but leaves IP risk with the user
Scalable’s current Content and Image Rights terms say generated images may carry machine-readable AI-generation markers and that users must not remove, alter or obscure them. The platform links the rule to EU AI Act Article 50. At the same time, Scalable says generated images are provided without general intellectual-property indemnity and does not warrant that outputs are free of third-party rights, except where enterprise customers separately negotiate protections in writing.
Why visual creators should care
Provenance and legal clearance solve different problems. A file can be correctly marked as AI-generated while still containing a trademark, copyrighted design or recognisable person that creates commercial risk.
Practical takeaway
Preserve AI markers, but do not treat them as rights clearance. Keep a separate approval step for copyright, trademark, likeness and client-specific restrictions before publishing or delivering generated imagery.
What to watch next
Watch whether provenance-bearing commercial image platforms increasingly pair machine-readable disclosure with optional human rights review or indemnification.
Sources
