Canva is putting more licensed Getty photography and video behind its AI development
Canva has expanded its long-running content partnership with Getty Images to support development of its generative-AI tools. Canva says the agreement forms part of its $200 million Content Fund and argues that creators and intellectual-property rights holders should receive royalties when their work is licensed for AI training. The company already operates a separate compensation programme for Canva Creators who consent to their content being used to train proprietary Canva models.
Why visual creators should care
The deal represents a very different training model from scraping publicly accessible images without direct licensing. For photographers and visual libraries, the crucial questions become what material is included, what rights Getty has granted, how contributors participate economically and whether training licences remain distinct from ordinary stock licensing.
Practical takeaway
Contributors should distinguish ordinary image licensing from AI-training licensing when reviewing agency agreements. Keep copies of the contributor terms that applied when work entered a collection and identify whether training revenue is shared separately from conventional downloads.
What to watch next
Watch whether Getty explains how participating contributors are compensated, what parts of its collection enter Canva’s training systems and whether photographers receive reporting that distinguishes AI-training revenue from ordinary licensing.
Sources
Meta’s new watermark shows how fragmented AI-image provenance is becoming
Meta has introduced Content Seal, an invisible watermark applied to images generated using its Muse model. Meta says the signal is designed to remain detectable after common transformations such as cropping, compression, resizing and screenshots. For now, verification is available through a separate testing tool, applies only to newer Muse images and does not yet cover generated video.
Why visual creators should care
Visual creators and publishers are increasingly facing several independent provenance systems at once: Content Credentials, SynthID and now Meta’s proprietary Content Seal. A signal that works only inside one company’s ecosystem is much less useful when images move through clients, publishers and competing social platforms.
Practical takeaway
Do not describe an image as verified simply because one proprietary detector returns a result. Record which provenance system was checked, which platform generated the asset and whether another independent signal such as C2PA is also available.
What to watch next
Watch whether Meta makes Content Seal detectable directly inside Facebook and Instagram, whether other platforms support it and whether Meta improves reliability after tests showed weaknesses when images were cropped.
Sources
Australian police are applying existing intimate-image laws to AI-generated likeness abuse
Queensland police have charged a 52-year-old man with 43 offences after investigating AI-generated posters allegedly targeting eight women. Police say the posters placed the women’s faces onto explicit sexual imagery and included names, social-media details and employment information. Charges include 27 counts of distributing intimate images and six counts of unlawful stalking. The allegations have not yet been proved in court.
Why visual creators should care
The case shows that synthetic imagery does not necessarily require a completely new offence before law enforcement can act. Existing intimate-image, stalking and communications laws can apply when identifiable real people are inserted into fabricated visual material.
Practical takeaway
Anyone responding to synthetic likeness abuse should preserve both the generated image and its publication context. Record URLs, physical locations, account information, dates, screenshots and the genuine source photographs that may have been used to construct the image.
What to watch next
The accused is due before Brisbane Magistrates Court on August 19. Watch how prosecutors characterise the AI-generated material and whether courts treat fabricated imagery in the same way as conventional intimate-image distribution.
Sources
Google is turning verified likenesses into reusable professional video assets
Google Vids now allows eligible users to create synthetic video clips using a personal avatar built from verified face and voice recordings. The avatar is managed through the user’s Google Account and can be selected as a character in Gemini Omni generations. Workspace administrators can disable the feature at domain level. Google says the underlying face and voice recordings are not used to train its generative models except for abuse and safety purposes.
Why visual creators should care
A digital likeness is becoming something that can sit inside an organisation’s normal production software. That makes synthetic identity an asset-management issue as well as a consent issue, particularly when employees, presenters or clients can appear in generated material without a new physical recording session.
Practical takeaway
Organisations should define who may create a personal avatar, where it may be used and whether generated performances require separate approval. A person consenting to create an avatar should not automatically mean every future commercial use is authorised.
What to watch next
Watch whether organisations develop internal likeness policies, how avatar rights are handled when an employee leaves and whether Google introduces project-level controls over who may use an approved digital identity.
Sources
YouTube is generating thumbnails tailored to a creator’s own visual style
YouTube has added thumbnail generation to Ask Studio for long-form videos. Creators can ask Studio to create a thumbnail for a specific video and then alter colours, layout and other elements conversationally. YouTube says the generated thumbnails are tailored to the video’s themes and to the creator’s specific style.
Why visual creators should care
Thumbnail design is a significant visual-creative market in its own right. Moving generation directly into the publishing platform reduces the distance between a creator’s existing visual identity and automatically generated promotional imagery, while potentially reducing demand for externally commissioned thumbnail design.
Practical takeaway
Creators and agencies should preserve the final approved thumbnail separately from automatically generated suggestions and record when AI generation was used. Where brand assets, photography or a recognisable design system are involved, decide who owns the resulting promotional asset and whether it may be reused outside YouTube.
What to watch next
Watch whether YouTube explains what material it uses to infer a creator’s 'specific style', whether generated thumbnails receive AI provenance information and whether agencies can disable or control generation for managed channels.
Sources
For AI-assisted visual work, documenting human contribution is becoming part of ownership
A Reuters Legal analysis of generative AI in game development describes concept art and other creative assets as part of a wider shift towards hybrid production. The analysis argues that copyright still depends on meaningful human authorship and that records showing creative direction, iterative development, editing and human oversight can support ownership and provenance claims. It also stresses the importance of retaining the AI-service terms that applied when an asset was created.
Why visual creators should care
The principle extends directly to illustration, concept art, design and AI-assisted image making. Simply saying that a person prompted an image generator may provide weak evidence of authorship. A documented sequence of sketches, selections, revisions, compositing and editing provides a much stronger account of human creative control.
Practical takeaway
Preserve intermediate visual stages rather than only prompts and final exports. Record reference gathering, sketches, generated alternatives, selections, manual edits, model versions and the terms of the service used at the time.
What to watch next
Watch whether clients, publishers and registration authorities increasingly request process evidence when determining ownership of AI-assisted commercial artwork.
Sources
