Federal prosecutors are beginning to use the TAKE IT DOWN Act against explicit AI replicas
US federal prosecutors have charged two men accused of creating and publishing sexually explicit AI-generated images and videos depicting female celebrities and private individuals. Prosecutors say the material drew millions of views. The defendants are among the earliest people charged under the TAKE IT DOWN Act and face potential prison sentences of up to two years.
Why visual creators should care
The cases move synthetic-likeness protection beyond platform reporting and civil takedown requests. They show that publishing an unauthorised explicit replica can create direct criminal exposure even when the depicted event never occurred and the image was generated rather than photographed.
Practical takeaway
People representing models, performers or other subjects should preserve the original source photograph, generated output, URLs, account names, publication dates and evidence connecting the replica to the depicted person. Report harmful material promptly, but retain copies before removal.
What to watch next
Watch how prosecutors prove identity, lack of consent and publication responsibility; whether altered crops and reposts are included; and how courts distinguish the generator, uploader, platform operator and commercial beneficiary.
Sources
Cyprus has created a new right over synthetic imitations of personal identity
Cyprus amended its copyright legislation in March to introduce a new sui generis right against unauthorised deepfake imitations. The provisions cover public distribution of synthetic imitations of a performer’s performance and of a natural person’s physical characteristics, including morphology, voice, biometric traits and behavioural features. Protection for personal characteristics continues for 50 years after death.
Why visual creators should care
The law does not limit protection to celebrities, commercial endorsements or explicit imagery. It gives ordinary individuals and performers a distinct right against synthetic identity imitation and uses copyright-style civil and criminal enforcement mechanisms.
Practical takeaway
Model and performer agreements involving AI should define face, body, voice, movement, behavioural characteristics and reusable digital replicas separately. A conventional release permitting publication of a photograph should not automatically authorise the creation of new synthetic appearances or performances.
What to watch next
Watch how Cypriot courts interpret express consent, whether the right can be licensed or transferred, how parody and public-interest exceptions operate and whether other EU countries adopt comparable copyright-style protections.
Sources
Content Credentials now support more complex professional production records
Version 2.4 of the C2PA specification adds new asset-format support, assertions and a JSON-based representation of Content Credential data. It also develops support for live video, cryptography, repository receipts and validation reporting. A repository-receipt assertion can record that a manifest has been ingested by a recognised credential repository.
Why visual creators should care
Professional visual production increasingly involves moving files between cameras, editing systems, agencies, asset managers, broadcasters and publishing platforms. Repository records and improved live-video support can help preserve evidence when provenance data is no longer carried only inside one static image file.
Practical takeaway
Studios adopting Content Credentials should test the full delivery chain rather than verifying only the original export. Record whether manifests survive conversion, asset-management ingestion, video packaging, client downloads and publication.
What to watch next
Watch which camera, video, DAM and publishing systems implement version 2.4; how repository receipts are displayed to users; and whether the JSON view improves auditing without being mistaken for independently verifiable provenance.
Sources
An image can carry two valid authenticity signals that make opposite claims
Researchers have demonstrated what they call an integrity clash: an image can contain a cryptographically valid C2PA manifest indicating human authorship while its pixels carry a valid watermark identifying it as AI-generated. Both systems may pass their own verification checks because provenance metadata and watermark detection operate independently.
Why visual creators should care
Agencies, competitions and publishers may assume that using two authenticity systems provides confirmation. The research shows that separate systems can instead produce a formally valid contradiction, leaving a reviewer to decide which production claim is credible.
Practical takeaway
Do not inspect Content Credentials and AI watermarks as unrelated pass-or-fail tests. Record both results together and escalate contradictory signals for source-file, signer and workflow review before accepting or rejecting the image.
What to watch next
Watch whether C2PA implementations and watermark providers introduce cross-layer checks, whether verification interfaces display conflicts clearly and whether platforms preserve enough information to investigate how the contradiction arose.
Sources
Photographers are building a human-made market outside social-media algorithms
The new 5 Boro Foto Fest in New York will create an in-person market for photography prints, books, zines and mixed-media work. AI-generated artwork is prohibited. The organisers describe the event as an attempt to connect photographers directly with audiences and build a small creative economy around work made by identifiable human artists.
Why visual creators should care
Visual creators are not responding to synthetic competition only through law, labels and platform policies. Some are creating separate markets in which human authorship, physical objects and direct contact with the photographer are part of the value offered to buyers.
Practical takeaway
Photographers selling human-made work should make the production claim specific and supportable. Explain the capture and printing process, retain originals and edition records, and avoid vague authenticity language that cannot be verified.
What to watch next
Watch whether collectors respond to human-made guarantees, whether other festivals and marketplaces adopt similar exclusions and whether provenance records become part of print and photobook sales.
Sources
A commercial concept-art workflow shows why AI use needs feature-level records
Concept artist Daniel Porto describes a hybrid commercial workflow that begins with human research, mood boards, drawing and photobashing. AI is introduced after the creative direction is established, using generated components, ControlNet and inpainting. The final work returns to conventional compositing, retouching, colour grading and detailed correction.
Why visual creators should care
A finished image described simply as AI-generated would conceal most of the production process. The example shows why professional records should distinguish the artist’s source draft, reference gathering, controlled generation, asset creation and final human finishing.
Practical takeaway
Record the workflow in stages. Preserve the initial sketch or composite, references, generated components, model and control settings, intermediate selections and the layered final file. This creates stronger evidence of human direction than a prompt log alone.
What to watch next
Watch whether agencies begin requesting feature-level AI disclosures, whether clients value documented human control and how contracts allocate rights in generated components combined with conventional artwork.
Sources
