AI and training data - US cases track Singapore law in preventing piracy claims over AI training data

One of the main questions about training AI Large Language Models (LLMs) is whether an AI company can use vast amounts of copyrighted data to train the LLM. In the US a number of authors brought a case against Open AI arguing their works had been reproduced to train ChatGPT. The courts distinguished inputs and outputs, saying that training (or 'digestion') was not infringement unless the output from the AI amounted to reproduction - click link here.

in 2021 Singapore legislated something similar - a specific copyright Infringement defence in the Copyright Act for AI machine learning called the Computational Analysis defence - for full details click here: Artificial Intelligence in Singapore. Singapore's aim in moving quickly on this has been to attract AI companies to the country safe in the knowledge that they will not be sued. Singapore's law and the US cases are nicely aligned and should set the global benchmark.