Meta has won an early round in the 3D-artist AI-training case
US District Judge Charles Breyer has dismissed 3D artist Austin Beaulier’s proposed class-action claims accusing Meta of violating the Digital Millennium Copyright Act by removing copyright-management information from 3D models allegedly used in generative-AI training. According to the August 26 ruling, Beaulier had not plausibly alleged the required intentional removal or the external distribution element relied on by his DMCA theory. The dismissal is not necessarily the end of the claim: he has 21 days to amend.
Why visual creators should care
Visual creators often assume that missing credit or stripped metadata automatically creates a strong DMCA claim. The ruling shows that proving removal, knowledge and the relevant downstream use can be much harder, especially when copyrighted assets are copied internally as part of a training pipeline.
Practical takeaway
Preserve more than the licence attached to an online asset. Keep evidence of the original copyright-management information, where the file was published, how it was obtained and what happened to that information later. A missing credit by itself may not prove the legal elements of a DMCA claim.
What to watch next
Watch whether Beaulier files an amended complaint within 21 days and whether he can plead more specific facts showing intentional removal and actionable distribution.
Sources
Talent agencies and SAG-AFTRA are closing ranks around explicit permission for AI replicas
SAG-AFTRA, Cameo, CAA, UTA, WME and the Music Artist Coalition issued a joint statement on August 26 supporting stronger protection for performers’ names, voices, likenesses and creative works in AI-generated content. Cameo says talent retain ownership of the videos they submit, that its policies prohibit users from making deepfake videos of talent without explicit authorization, and that Cameo does not use talent videos to create or train AI digital replicas without permission.
Why visual creators should care
The statement brings unions, agencies and a commercial talent platform around the same basic principle: access to a performer’s existing video is not permission to generate a digital double from it.
Practical takeaway
For photography, advertising and film work, add synthetic-replica permission as its own line in talent agreements. A conventional release covering photographs or filmed performance should not silently become permission to train or generate a reusable AI likeness.
What to watch next
Watch whether the coalition develops common contractual language or technical controls for digital-replica authorization and whether other talent platforms adopt similar no-training commitments.
Sources
Higgsfield’s new terms make account type a training-rights decision
Higgsfield’s revised Terms of Use become effective on August 27 for users who registered before July 26 and had not accepted them earlier. Under the general terms, Higgsfield says user content, inputs and outputs may be used to train and improve its AI systems. Users can stop future use by deleting their content or account, but material already incorporated into trained models cannot feasibly be removed. Higgsfield separately says enterprise and business customers operating under an Enterprise Agreement receive different treatment: their content is not used to train or improve Higgsfield models and is handled as confidential.
Why visual creators should care
Two creators can use the same visual platform and have materially different training protections because one is using ordinary consumer terms and the other has a business agreement.
Practical takeaway
Record the account type and governing agreement for client production. If a client requires no-training treatment, confirm that the actual workspace qualifies rather than relying on the platform name alone.
What to watch next
Watch whether Higgsfield introduces a consumer-level training opt-out and whether the enterprise no-training distinction becomes visible directly inside account settings.
Sources
Google is bringing visible AI disclosure into the advertising workflow
Google’s current Display & Video 360 product roadmap says advertisers will be able to disclose when creative assets were generated or edited using AI. In applicable regions, including markets governed by transparency rules in the EU, India and New York, the disclosure will appear as a visible AI label. Google also says the information will appear globally inside the About This Ad panel.
Why visual creators should care
For agencies, AI disclosure is becoming part of media trafficking rather than a note added manually to a creative file. That makes whoever uploads and configures the campaign part of the disclosure chain.
Practical takeaway
Add AI status to the handoff between creative and media teams. The person booking the ad may need to know whether an image was captured, materially AI-edited or generated even if that distinction is obvious to the designer who created it.
What to watch next
Watch the rollout date and the exact threshold Google uses for 'edited using AI', particularly for commonplace AI-assisted retouching and automated image optimisation.
Sources
Synthesia’s AI press officer shows what an authorised synthetic representative can look like
Synthesia launched an Interactive Avatar on August 26 that acts as an AI press officer for the company. It uses the voice and likeness of Synthesia’s own head of corporate affairs, Alexandru Voica, and answers questions in real time using company information supplied through a language model. The page visibly states that the content was generated using AI and is labelled in accordance with EU AI Act Article 50 requirements.
Why visual creators should care
Most likeness stories involve unauthorized replicas. This is the opposite production model: a named individual knowingly supplies the face and voice, the synthetic role is clearly defined, and the audience is told that the interaction is AI-generated.
Practical takeaway
For synthetic spokespeople, record who authorised the replica, what role it may perform, what source information it may use and how viewers will be told that they are interacting with AI.
What to watch next
Watch whether companies begin using authorized interactive replicas for customer service, sales and corporate communications and whether contracts start defining what the replica may say on the person’s behalf.
Sources
Content Credentials are starting to sit alongside ordinary editorial photo metadata
Current photographs distributed through the US Defense Visual Information Distribution Service show a Content Credentials section alongside conventional editorial information including photographer, capture date, location, resolution and image identifier. Examples published August 26 include official US Navy photographs made aboard USS Nimitz and USS James E. Williams.
Why visual creators should care
Provenance becomes much more useful when it is part of an ordinary picture-desk workflow rather than something a viewer has to inspect with a specialist tool. The DVIDS pages put credential information next to the same metadata editors already use to judge source and attribution.
Practical takeaway
If you supply editorial or documentary photography with Content Credentials, make sure the publishing system exposes the provenance rather than merely preserving it invisibly inside the file.
What to watch next
Watch whether news agencies, government image libraries and editorial CMS platforms adopt similar reader-facing presentation of Content Credentials.
Sources
