Runway draws a clearer line around the people and artists you can put into AI video
Runway’s current usage policy prohibits using another person’s image, video or audio without permission. It also prohibits content that may violate intellectual-property rights and attempts to create material in the style of a known living artist. Runway says automated systems and human review are used to enforce the policy and accounts can be suspended for violations.
Why visual creators should care
AI video increasingly starts with existing photographs, footage, voices and visual references. Runway is making the uploader responsible for having permission before those materials enter the generation process rather than treating rights clearance as something to consider only when the finished video is published.
Practical takeaway
For client or performer work, record permission to upload a face, voice, photograph or video into a generative system separately from permission to publish the original material. Do not assume an ordinary model release automatically covers synthetic performance generation.
What to watch next
Watch whether Runway provides more granular consent tools for professional productions and how it interprets the restriction on known living artists as image and video reference workflows become more sophisticated.
Sources
One film festival is defining human-made work by banning generative AI completely
The Sparknify Human vs. AI Film Festival places filmmakers into separate human-produced and AI-generated categories. Human entries must use conventional cameras and authentic human voice recordings, and generative AI is prohibited at every stage of production. The AI category takes the opposite approach: the entire film must be made using generative AI, with conventional software limited mainly to post-production assembly, colour correction and sound mixing.
Why visual creators should care
The festival avoids the increasingly difficult question of how much AI is too much by defining two production processes rather than trying to judge every hybrid workflow individually. That gives filmmakers a clear standard but also shows how difficult the middle ground has become.
Practical takeaway
If a commission, award or festival distinguishes human and generated production, keep enough evidence to prove the category. Camera originals, voice recordings, project files and generation histories may matter more than the appearance of the finished film.
What to watch next
Watch whether more film and photography institutions create separate production categories or whether a third AI-assisted category emerges between fully captured and fully generated work.
Sources
Reply wants the AI production record as well as the finished film
The Reply AI Film Festival requires submitted films to be generated or enhanced through AI. Entrants may combine AI with conventional filmmaking, but they must submit a separate document explaining how AI was used in the production. The festival explicitly recognises AI use across scripts, storyboards, images, video, sound and visual effects.
Why visual creators should care
This is a useful shift from disclosure as a yes-or-no label to disclosure as a production description. Two films may both be called AI-assisted while using the technology in completely different ways.
Practical takeaway
Keep an AI-use note while the production is happening rather than reconstructing it afterwards. Record the tool, model, source material, generated element and human decision for each material use of AI.
What to watch next
Watch whether commercial clients, insurers and other festivals begin requesting similar AI production statements rather than relying on a generic AI disclosure.
Sources
Another film competition is judging the responsibility of the AI use rather than the presence of AI itself
California Community Colleges’ Reel Futures competition closed submissions on August 14. Films could be made with or without AI tools as long as they meaningfully engaged with AI as a theme or context. Where AI tools were used, effective and responsible AI use formed part of the judging criteria, while all submitted material had to be original or properly licensed.
Why visual creators should care
This provides a different model from either banning AI or creating a separate synthetic category. The production remains primarily the filmmaker’s work, while AI use becomes something that can be evaluated for responsibility and creative purpose.
Practical takeaway
When AI forms only one part of a project, record why it was used and what it contributed. That is often more useful than simply labelling the entire film 'AI-assisted'.
What to watch next
Watch whether professional competitions begin judging AI use in a similar way, particularly where copyright clearance, source assets and the human creative role can be demonstrated.
Sources
The Royal Society keeps its photography competition on the camera side of the line
The Royal Society's photography competition rules require submitted images to have been taken with a camera and state that AI-generated images cannot be considered. The rule creates a straightforward eligibility boundary based on the origin of the image rather than attempting to distinguish good and bad uses of generative technology.
Why visual creators should care
Scientific and documentary photography depend heavily on the relationship between an image and something that existed in front of a camera. As synthetic imagery becomes more convincing, institutions increasingly need to decide whether that relationship is part of the definition of photography itself.
Practical takeaway
For competitions that require camera-made work, preserve the original capture and read the editing rules separately. A prohibition on generated images does not necessarily tell you which AI-assisted adjustments are permitted in post-production.
What to watch next
Watch whether science and documentary competitions begin requiring RAW files, Content Credentials or edit histories alongside simple bans on generated imagery.
Sources
A visual glitch in a deepfake is not the same thing as protection from one
Recent reporting has examined claims that tattoos can make people harder to reproduce convincingly in AI-generated video because current systems may struggle to keep complex body details consistent across frames. Experts quoted in the report caution that this is a temporary technical weakness rather than reliable protection. Still-image generators are already less affected and video consistency continues to improve.
Why visual creators should care
People increasingly look for visible features that might make synthetic copying difficult, but those defences depend on the limitations of today's models. A photographer, performer or creator should not assume a distinctive tattoo, hairstyle or other visual detail meaningfully prevents their likeness from being recreated.
Practical takeaway
Treat likeness protection as a rights and evidence problem rather than a visual-obfuscation problem. Preserve original images and publication history, restrict synthetic reuse contractually where possible and use platform reporting or detection tools when unauthorised replicas appear.
What to watch next
Watch how quickly video models improve temporal consistency around tattoos, jewellery and other persistent identity details, and whether provenance or likeness-matching systems become more useful than visual countermeasures.
Sources
