Europe’s biggest pending AI copyright opinion has slipped without a new date
Advocate General Maciej Szpunar had been expected to deliver his opinion in Like Company v Google Ireland on September 3, but the opinion was not published. The case is not listed for delivery in the Court of Justice calendar for September 3–10 and no replacement date has been announced. The Grand Chamber case asks, among other things, whether training a generative model on protected material amounts to reproduction under EU copyright law and, if so, whether that use can fall within the Article 4 text-and-data-mining exception. The final judgment will follow separately and the Advocate General’s opinion will not itself bind the Court.
Why visual creators should care
Although the underlying dispute involves press publications, the interpretation of reproduction and the EU text-and-data-mining exception could affect far more than publishers. Photographers, illustrators and other visual rights holders rely on the same copyright framework when reserving works against commercial AI training.
Practical takeaway
Do not treat September 3 as having produced a new EU rule. Nothing has been decided. For visual work published online, continue to preserve copyright records and use machine-readable training reservations where the EU text-and-data-mining opt-out is part of your rights strategy.
What to watch next
Watch the CJEU calendar for a replacement date and Germany’s separate Federal Court ruling in the photographer-versus-LAION case, which may arrive before Luxembourg provides its broader EU signal.
Sources
European visual creators have only days left to feed into the Commission’s next AI strategy
The European Commission’s call for evidence on a future AI strategy for the Cultural and Creative Sectors remains open until September 11. The Commission expressly invites artists, creators, rights holders, professional associations and other stakeholders to contribute. Its stated aim is to develop a strategy in which AI complements rather than replaces human creativity while supporting a fair and sustainable creative economy. The initiative will also be aligned with the Commission’s ongoing copyright work.
Why visual creators should care
Visual creators often encounter AI policy after the rules have already been drafted. This consultation is an opportunity for photographers, illustrators, filmmakers, designers and their organisations to put training consent, attribution, licensing, provenance and human creative value into the policy record before the strategy is written.
Practical takeaway
Creator organisations with a position on AI training, licensing or replacement of professional visual work have until September 11 to submit it to the Commission rather than waiting for the eventual strategy.
What to watch next
Watch the Commission’s resulting strategy for specific measures around licensing, creator remuneration, training transparency, provenance and adoption of AI within cultural and creative businesses.
Sources
Freepik shows why reference-image permission, output ownership and AI training need separate records
Freepik’s current AI Product Terms cover text, image, video and other inputs used with its generative tools. Users warrant that they hold the necessary rights to uploaded material and have permission from people whose image, voice or identity appears in it. Paid subscribers receive Freepik’s rights in generated output to the fullest extent permitted by law, although Freepik warns that AI outputs may not qualify for copyright protection in every jurisdiction. The company also says it does not use users’ inputs or outputs to train its own AI models.
Why visual creators should care
The platform separates three questions that are frequently collapsed into one: whether you had permission to upload a photograph, what rights the platform gives you in the result and whether your material can train the provider’s models.
Practical takeaway
For professional generations based on client or model photography, save the source-image licence or release alongside the generation record. Output ownership cannot cure an unauthorised input.
What to watch next
Watch third-party models available through Freepik, because its terms incorporate provider-specific requirements and those can create additional restrictions beyond the Freepik account itself.
Sources
Magnific lets you commercialise an AI image without giving it the legal protection attached to stock
Magnific’s current AI copyright guidance says paid users may use generated images and videos commercially, including as central elements of professional projects. But AI-generated content is expressly excluded from the standard legal protection applied to licensed stock assets. Magnific also warns that it cannot guarantee copyright ownership of generated output in every jurisdiction and leaves responsibility for third-party rights with the user. Enterprise agreements may separately provide enhanced protection or indemnification.
Why visual creators should care
The distinction is important for designers and agencies who use both stock and AI generation inside the same visual platform. A conventional licensed photograph can arrive with contractual protection that a visually similar generated image does not.
Practical takeaway
Record whether an asset was licensed from the stock library or generated with the AI tool. They can carry materially different legal protection even when both appear inside the same subscription.
What to watch next
Watch whether indemnity for generated imagery moves from enterprise-only agreements into mainstream professional plans as visual AI becomes a larger part of commercial production.
Sources
Shutterstock is using human review as the gateway to legal protection for AI images
Shutterstock’s current guidance says whether a generated image is indemnified depends on the customer’s plan and whether the image has passed Shutterstock’s human content review. Where a plan includes that review and the generated image is approved, Shutterstock says the image receives its representations, warranties and indemnity protection for uses including advertising and client delivery. Unreviewed generated images do not receive that protection.
Why visual creators should care
This turns human inspection into part of the commercial rights workflow rather than merely a quality-control step. Two images made with the same generator can therefore carry different contractual risk depending on what happened after generation.
Practical takeaway
If a client requires indemnified AI imagery, preserve evidence that the exact delivered asset passed the relevant human review. Do not assume every generation made through a stock platform receives the same legal cover.
What to watch next
Watch what Shutterstock’s reviewers assess for likenesses, logos, artworks and other potentially protected visual elements and whether review status becomes embedded in downloadable licence records.
Sources
Unsplash lets developers display photography through its API but not turn the API into a training dataset
Unsplash’s current API guidance requires high-quality, authentic uses of its photography and expressly tells developers to stay away from data mining and training AI models. The restriction applies to use of the API rather than changing the underlying copyright ownership of the photographs.
Why visual creators should care
Access and training permission are different things. A developer may be authorised to retrieve and display photographs through an API while remaining contractually prohibited from converting that feed into training data.
Practical takeaway
When photographs enter an AI project through a stock service or API, check the acquisition route as well as the visible image licence. Technical access does not itself establish training permission.
What to watch next
Watch whether more photography services attach machine-readable AI-training restrictions directly to API responses and whether platforms begin auditing developers for dataset-building behaviour.
Sources
