AI Image and Video Generators vs Copyright: What UK Creators Need to Know in 2025

“You Can’t Escape Censorship”: What a Reddit Video Experiment Says About AI, Copyright and Filters in 2025

A Reddit creator set out to “test the limits” of modern image and video models and ran straight into tightening content filters. They used multiple tools – Nano Banana Pro, Kling and Seedance – via different platforms and APIs. Midway through the project, the user says restrictions were tightened, to the point where anything that “remotely looked like Mickey” was blocked by Seedance 2.0.

“This is an unauthorised artistic creation produced for the purpose of social critique and the defence of creative freedom… a non-commercial, independent art piece.”

That tension – creative experimentation vs platform rules and intellectual property (IP) risks – is exactly where many UK creators and developers now work. Here’s what the post means, why platforms are getting stricter, and what UK law really says about parody, characters and reusing protected material.

Read the original Reddit thread.

What actually happened in the Reddit test

The poster describes producing a short video over a few days, largely with image-generation tools, and layering outputs to make something “enjoyable to create”. During that period:

• Safety policies reportedly tightened; prompts or images that looked “Mickey-like” began to trigger blocks in Seedance 2.0.
• The creator framed the work as non-commercial parody and social critique, including a disclaimer about metaphorical violence.

We don’t get technical details, model versions or prompts – not disclosed. But the pattern will feel familiar to anyone using hosted generators: policy shifts can land abruptly, and character/trademark filters often trip even borderline cases.

Why image and video platforms are tightening copyright filters

Three overlapping reasons explain the trend:

• Legal exposure: Character likenesses and brand assets attract takedowns and claims. Hosts don’t want to arbitrate borderline “parody” calls at scale.
• Alignment policies: Safety teams train filters to block named styles, celebrities, protected characters and violence. These filters iterate frequently, sometimes aggressively.
• Commercial relationships: Market access, app-store policies and ad partners all nudge platforms to reduce IP and safety risk.

Developers should expect more false positives, broader “style” blocks and quicker policy updates – especially around iconic characters and logos.

UK copyright and parody in 2025: what’s actually allowed?

Under UK law (Copyright, Designs and Patents Act 1988), “fair dealing” exceptions allow limited use of copyrighted material for specific purposes, including caricature, parody or pastiche. Key points:

• It must be fair dealing: proportionate, necessary for the purpose, and not a substitute for the original.
• Commercial vs non-commercial: non-commercial helps but is not decisive.
• Moral rights: UK authors may have rights to be identified and to object to derogatory treatment of their work.
• Separate rights still apply: trademarks, passing off, privacy/data issues and platform contracts can still bite.

Parody is not a free pass. If your work trades on consumer confusion, includes logos, or harms a brand’s market, you face risks regardless of “fair dealing” arguments.

Useful overview: GOV.UK – Exceptions to copyright (including caricature, parody and pastiche).

Mickey, trademarks and public domain: why this stays tricky

There’s extra confusion around famous characters. In the US, the earliest “Steamboat Willie” incarnation of Mickey Mouse entered the public domain in 2024. That does not grant a blanket right to use modern Mickey visuals, and it does not automatically apply in the UK in the same way.

• Jurisdiction differs: public domain status can vary by country and by version of a character.
• Trademarks endure: Disney’s trade marks over names, logos and character indicators remain. Even if a depiction is PD, trade mark law can restrict uses that cause confusion or imply endorsement.
• Platform rules go further: hosts often block “Mickey-like” outputs entirely to avoid grey areas, regardless of legal nuance.

If your work even glances off iconic characters, expect automated filters to intervene, and plan for manual review or alternative creative choices.

More on trade marks: GOV.UK – Trade marks.

Who owns AI-generated outputs in the UK?

Ownership of AI outputs is unsettled globally. The UK is unusual in recognising “computer-generated works” where there is no human author; in that case, the “author” is the person who made the arrangements necessary for creation, with a 50-year term. In practice, platforms’ terms of service and your degree of human input matter a lot, and disputes are rising.

For commercial projects, get contracts straight: who owns the prompts, the outputs, and the right to exploit them? Hosted services may reserve rights or restrict use cases.

Practical guidance for UK creators using image/video generators

• Make the parody obvious: build a clear transformative message or critique. The more it comments on the original, the better your fair dealing position.
• Avoid confusion: steer clear of logos, brand fonts and elements that imply endorsement. Use generic styles instead of named characters.
• Document your process: keep a prompt log, timestamps, and model versions to show intent and proportionate use. A simple Sheets-based log works well – here’s a workflow to connect models and Sheets for governance and auditing: How to connect ChatGPT and Google Sheets (Custom GPT).
• Check the platform’s licence: hosted tools may ban use of protected characters or violent content. Even lawful parody can be blocked by contract.
• Consider open/local models: if you have strong compliance processes and legitimate parody grounds, local models reduce platform-level blocks – but put the legal risk squarely on you.
• Manage biometric and privacy risk: avoid real people’s faces without consent. UK GDPR can apply if images are used to uniquely identify individuals, and passing off/defamation are separate risks.
• Expect takedowns: have a plan to respond with fair dealing arguments, process notes and alternative edits. Non-commercial intent helps, but isn’t decisive.

Implications for UK developers and businesses

• Filters will tighten: budget for more iteration, higher rejection rates and lag from moderation pipelines. Build retries and fallbacks into your tools.
• Policy shifts mid-flight: product roadmaps need “policy volatility” buffers. Keep a changelog of model and platform updates.
• Governance by design: add prompt whitelists, named-entity redaction and style safety layers before hitting the vendor API.
• Clear rights chain: if you ship media to clients, document model sources, training disclosures (if available), and usage terms.
• When in doubt, commission: for brand-critical assets, a licensed illustrator or animator can be cheaper than fighting a takedown.

Bottom line: creative freedom meets automated risk management

The Reddit post captures a real, growing dynamic: even good-faith parody can be throttled by automated filters, especially around iconic characters. In UK law, parody and pastiche exist – but they’re bounded by fairness, and they don’t override trademarks or platform contracts.

If you’re experimenting, build with intent, document your process, and avoid confusion with protected brands. If you’re commercial, invest in governance and clearance. Either way, assume the filters will keep getting stricter, not looser.

Further reading

• Original Reddit post
• GOV.UK – Exceptions to copyright (including parody)
• GOV.UK – Trade marks overview

This article is general information for UK readers, not legal advice.