BFI Report: 130,000 Copyrighted Scripts Used to Train AI Models
The British Film Institute reported that generative AI models have been trained on 130,000 copyrighted film and TV scripts without permission or payment to rights holders. The report warns this poses a direct threat to the UK screen sector through potential job losses and erosion of traditional business models, and calls for an opt-in regime for AI companies seeking to use copyrighted content.
What happened
Artificial intelligence companies have used 130,000 film and TV scripts to train generative AI models without payment to or permission from rights holders, according to a report by the British Film Institute.
The BFI report, published in partnership with Goldsmiths, Loughborough and Edinburgh universities, identifies the use of copyrighted intellectual property to train AI models as the 'primary issue' facing the UK's Β£125 billion screen sector.
Reported impact
The BFI warns that AI automation of tasks such as writing, translation and some technical visual effects and character animation could eliminate entry-level jobs that bring new generations of workers into the screen industries.
The report raises concerns that AI could erode traditional business models, displace skilled workers, and undermine public trust in screen content.
Index analysis
The UK creative industries are advocating for an 'opt-in' regime that would force AI companies to seek permission and strike licensing deals before using content, rather than the current opt-out model.
The report warns of a 'critical shortfall' in AI training provision in the UK screen sector, with most education being 'informal' rather than 'formal', leaving many workers, particularly freelancers, lacking resources to develop skills complementary to AI.
Governance controls
The UK government is currently in the process of considering what legislation to put in place to address the use of copyrighted material in AI training.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.