AI, Creativity, and Culture in 2026

In November 2025, Warner Music Group settled its copyright case against Suno, the AI music generator, with reports describing a multi-million-dollar payment plus a forward licensing partnership — and Suno’s acquisition of Songkick from Warner as part of the deal. A month earlier, Universal Music Group had settled with Udio, securing a per-generation royalty and a content-identification requirement against UMG’s catalogue. Sony has settled with neither, and its fair-use cases against Suno in Massachusetts and Udio in the Southern District of New York are expected to produce a defining ruling by mid-to-late 2026. Eighteen months after the RIAA filed those landmark suits, the music industry has its first real AI economic model — and the legal floor is still moving.

The creative industries in 2026 are the most contested ground in the AI economy. Below is the honest picture across music, film and video, writing, and visual art.

Music: licensing has arrived, but the floor is still uneven#

Suno and Udio remain the two largest consumer-facing generative music tools. Both have shifted decisively from “fair use will save us” rhetoric into deal-making mode after the Warner and UMG settlements. The shape of the new economic model — at least where deals exist — is a per-generation royalty for catalogue-similar outputs, higher rates for commercial distribution, and a content-identification obligation on the AI platform.

What that does not solve: independent artists and smaller labels who are not party to any of these deals. Their material is still being used in training datasets, and the only practical recourse to date has been opt-out registers — patchy in coverage and slow to propagate.

Streaming platforms have started filtering. Spotify, Deezer, and others have removed batches of AI-generated tracks that triggered fraud heuristics — typically streams in suspicious patterns by accounts whose only catalogue was synthetic. The volume removed has been measured in tens of thousands of tracks, not millions, but the direction of travel is clear: platforms will police synthetic output where it threatens royalty integrity.

Film and video: the Sora moment, and what it actually changed#

OpenAI’s Sora 2 release in late 2025, followed by Google’s Veo 3, Runway Gen-4, and Pika 2.0, dragged generative video into broad public use. The output quality at the high end is now closer to professional B-roll than to the uncanny early demos. A handful of independent shorts shot with AI-augmented pipelines were screened at major festivals across late 2025 and early 2026.

Hollywood pickup has been measured rather than dramatic. The major studios are using generative video almost exclusively in pre-production — concept reels, previs, mood films — and in marketing rather than in finished theatrical product. The reason is upstream: union agreements and IP exposure.

SAG-AFTRA’s expanding AI guardrails#

SAG-AFTRA ratified the Interactive Media Agreement in July 2025, which sets consent and disclosure requirements for digital replicas and a higher minimum — reported as 7.5 times scale — for “real-time generation” use cases such as AI-voiced game characters. In February 2025 the union approved an updated contract with Telemundo Television Studios with comparable AI protections. In May 2025 the union reached a tentative agreement with Nickelodeon covering voice actors. The Commercials Contract requires that advertisers seek SAG-AFTRA permission before authorising any third party to use covered material to train an AI system. The main TV/Theatrical contract with the studios expires in June 2026 and AI is expected to be the dominant negotiating axis.

The writer’s strike aftermath#

The 2023 WGA agreement set a baseline that AI cannot be a “writer” in the credited sense and that studios cannot require writers to use AI tools. Those protections have held in 2024–2026 negotiations and have been broadly imitated in other guild contracts. Studios that have leaned on AI in writers’ rooms have done so quietly and at the margin — research, outline assistance, table-read summaries — not as a replacement for staff writers.

Stacked vinyl records beside a folded paper film reel and a microphone silhouette representing AI's reach across music and film

Visual art: copyright clarity has crept forward#

The US Copyright Office moved meaningfully in 2025. Its January 2025 guidance confirmed that works where human creative expression remains evident can be registered, even where AI tools were used in production. The Office also registered “A Single Piece of American Cheese” in January 2025 — the first visual artwork composed solely of AI-generated outputs, registered as a composite work on the basis of human-driven selection, arrangement, and coordination.

In May 2025 the Office published its analysis of training and fair use, concluding that AI developers who use copyrighted works to train models that generate expressive content competing with the originals are likely to fall outside fair use — a meaningful narrowing of the broadest claims from generative AI vendors. The Andersen v. Stability AI case has continued to move forward.

The platform landscape#

Adobe Firefly Image 4 remains the most defensible enterprise option because of Adobe’s IP indemnification — the company commits to defend enterprise customers against IP claims arising from Firefly outputs, on the basis of its licensed-and-public-domain training corpus. Runway Gen-4 has a narrower indemnification position. OpenAI’s Sora has been the most opaque on training data, and OpenAI has not matched Adobe’s enterprise indemnification.

Midjourney shipped v7 in early 2026 with materially better photoreal quality and improved text rendering. Stable Diffusion’s open ecosystem continues to evolve through Stability AI and a wider community of fine-tunes; the commercial position has weakened relative to Firefly and Midjourney for enterprise use, but it remains the dominant choice for independent creators and self-hosted pipelines. Canva’s Magic Studio features have absorbed generative tooling into mainstream marketing workflows at small-and-medium business scale.

Writing, journalism, and publishing#

The publishing industry has been more cautious than music and film. Most major news organisations have either licensing deals with model providers (the AP, Axel Springer, Financial Times, News Corp with OpenAI; Reuters and others with various providers) or active litigation (the New York Times case continues). The BBC publicly threatened legal action against Perplexity over content reuse in 2025 and has set out a clear internal AI policy: no AI-generated journalism without human editorial control, transparent labelling where AI assists, no use of AI to imitate identifiable real-world individuals without consent.

In trade publishing, the picture is less settled. Several large publishers have signed deals selling backlist access to AI training; authors have, in many cases, only learned of it after the fact, and the Authors Guild has been pushing for opt-in defaults and a clearer royalty share.

The cultural backlash: AI slop and trust erosion#

“AI slop” — the avalanche of low-quality, low-trust synthetic content flooding social feeds, ad networks, and search results — has become the defining cultural complaint of 2025–2026. The term is now used regularly by mainstream press. Reader trust in unsourced content has slipped. Sora 2’s release in September 2025 accelerated the conversation because realistic synthetic video raised the stakes from “annoying” to “dangerous.”

The response has been both technical and editorial. Cultural institutions are publishing policies. The BBC’s editorial guidelines on AI are publicly maintained. The Smithsonian has set a public-facing approach on generative AI use across its programmes and exhibitions. Major libraries and museums are working through provenance standards. The Content Authenticity Initiative, led by Adobe with partners including the BBC, Microsoft, and major camera manufacturers, has continued to gain adoption for cryptographic content credentials.

Stylised canvas on an easel with a paint brush and a tablet stylus crossed in front of it

Education and museums: institutions setting norms#

Cultural and educational institutions have started to publish their own AI policies in a way that is shaping audience expectations. The BBC’s editorial guidance on AI use is publicly maintained and updated. The Smithsonian has signalled a deliberate, consultative approach to generative AI across its programmes, exhibitions, and educational content — leaning toward transparency, attribution, and human curation. National libraries in several European countries have published their own positions on training-data use of digitised collections, with several insisting on opt-in rather than default consent. Major universities have rewritten plagiarism and authorship policies to handle generative AI use in coursework. None of these documents has the force of law, but together they are shaping what audiences and students now expect a credible institution to do — and that, in turn, sets the floor for what serious creative brands are willing to ship in their own work.

What this means for creative-industry leaders in 2026#

Five working assumptions worth holding.

First, the legal floor is still moving and will not stabilise before the Sony rulings and the 2026 SAG-AFTRA negotiation. Plan for a moving target, not a settled framework.

Second, indemnification is now a real procurement criterion. If you are deploying generative tooling at brand or studio scale, Adobe Firefly’s position is materially different from Sora’s, and that gap shows up in legal exposure.

Third, the per-generation licensing model emerging in music will likely propagate, in some form, into other media. Build commercial workflows that can attribute, log, and report generations.

Fourth, audiences are getting more sensitive to synthetic content, not less. Visible labelling and provenance metadata — Content Credentials, watermarks, disclosure — are becoming brand assets, not compliance overhead.

Fifth, the workflow that wins is the one that treats AI as a creative-direction tool — generation under taste and editorial control — rather than a creative-output replacement. Klarna’s customer-service reversal is the cautionary tale that applies here too: ship AI as an instrument, not a substitute.

Where pdpspectra fits#

Creative-industry clients usually need three things at once: a generative tooling stack they can defend legally, content-credentials and watermarking infrastructure, and an evaluation harness for the model outputs that ship to audiences. Our AI and LLM integration practice covers exactly that scope — picking the right model and provider, building the provenance pipeline, and the production guardrails that keep your brand on the right side of the moving legal floor.

If you are a studio, label, publisher, or brand thinking through a defensible 2026 generative-AI position — and want help that has actually shipped this work — we are happy to walk through your stack.