Momentum has not been lost following last week’s rallying cry at Westminster including Anti Copying In Design (ACID) who joined UK Music’s fantastic event to support all organisations and their members in design, publishing, authors, photographers, film makers, games, film, etc, the lifeblood of our creative economy,

The ever diligent and proactive Creative Coalition for AI, in its continuing battle for the creative economy to be protected, has produced some timely myth busters to explain some of the many misunderstandings around AI and IP, which underpin the need for continuing advocacy for creators’ rights

Please see the full Creative Rights in AI Coalition: Copyright and AI Mythbuster

Myth 1 – AI training is just like human learning from creative work ​

• AI training involves mass scraping of works, unlike human learning from individual books. ​
• Creators should not have to rent back their own work used in AI models. ​
• Generative AI models are prediction engines that rely on creative inputs, risking creators’ livelihoods. ​

Myth 2 – There is a lack of clarity in existing copyright law ​

• UK copyright law requires permission for commercial text data mining for AI training. ​
• Exceptions for non-commercial research and temporary copying do not apply to AI models.
• Lack of transparency over works used in AI training hinders creators’ ability to assert rights. ​

Myth 3 – It is not feasible for AI firms to agree with licences for content ​

• AI firms often do not attempt to license content before using it. ​
• Many AI firms have successfully licensed content from news publishers. ​
• A significant majority of artists are willing to license their work for AI training. ​

Myth 4 – Individual works have no real value in AI training ​

• The value of individual works in AI training datasets is significant, contrary to AI firms’ claims. ​
• High-quality data is particularly valuable during model fine-tuning and retrieval-augmented generation.
• UK creative content is valuable to AI firms, and they should be willing to pay for it. ​

Myth 5 – Opt-out will give creators control ​

• An effective opt-out mechanism does not currently exist, as acknowledged by the government. ​
• Opt-out places the burden on small creators to control their work against powerful tech firms. ​
• Existing opt-out tools are often ignored, leaving creators vulnerable to unlicensed scraping.

 ​

Myth 6 – We must mirror more permissive copyright regimes in other jurisdictions ​

• The UK should not adopt a more permissive copyright regime based on US or EU practices.
• Recent US court cases suggest a shift towards stronger copyright protections. ​
• The UK has a unique opportunity to create solutions that benefit both creative and tech sectors. ​

Myth 7 – Our national security will be put in danger if we retain strong copyright law ​

• Advanced AI models do not require creative content for key areas like national security and public services. ​
• Reliance on US Big Tech models poses a greater risk to national security. ​
• Quality journalism is essential for accurate AI outputs, which requires funding through copyright protections.

Myth 8 – Degrading copyright law will support AI investment ​

• Weaker copyright law primarily benefits US Big Tech firms, not UK investment. ​
• UK AI trade bodies oppose government proposals that would favour large tech companies. ​
• Energy costs and data protection laws are more significant factors for investment decisions. ​

Myth 9 – Transparency must be accompanied by opt-out ​

• Transparency provisions can exist independently of an opt-out mechanism. ​
• Lack of transparency, not law, hinders creators from asserting rights. ​
• Implementing transparency could reduce infringement risks for AI firms. ​

Myth 10 – Granular transparency is not feasible or would compromise trade secrets ​

• Granular transparency can be achieved through private disclosure obligations without public data dumps. ​
• AI models consist of billions of data points, making individual disclosures manageable. ​
• Transparency is crucial for addressing various AI policy issues, including online safety and bias prevention. ​

Please show your support by signing our ACID IP and AI Charter