Writers in Singapore resisted the Singapore government’s plan to train Southeast Asia’s first local large language model on their work due to copyright and compensation concerns.

The Singaporean government’s National Multimodal Large Language Model Programme (NMLP) sparked controversy among local writers and publishers. 

The NMLP, a SGD 70 million (USD 52 million) project launched in December 2023, aims to train a large language model (LLM) on locally produced material to address the bias of existing LLMs that are seen to have disproportionately large influences from Western societies.

Singapore’s government believes that an LLM trained on local material would have more accurate references to Singapore’s history, colloquialisms and culture, and would be able to understand widely spoken languages such as Malay, Mandarin and Tamil. 

However, the literary community, including well-known literary figure Gwee Li Sui, expressed concerns about the lack of clarity on how their works would be protected from being used for purposes other than cultural representation.

An email sent by the government initially gave respondents 10 days to respond to a survey asking them about the project, but contained few details on compensation or copyright protection. As a result, several writers, including Gwee, declined to let the LLM train on their works.

This resistance is part of a worldwide trend against the use of published works to train AI systems. 

For instance, US comedian Sarah Silverman and authors like John Grisham and George R.R. Martin have issued class-action lawsuits against OpenAI and Meta, accusing the companies of copyright infringement for using protected work to train AI programmes.