Google has submitted a report to the Australian government expressing concerns about the legal uncertainty and copyright issues that are hindering the development of AI research in the country. The company suggests that clarity is needed on potential liability for the misuse or abuse of both general-purpose and specialised AI systems.

Google also calls for the establishment of a new copyright system that allows for appropriate and fair use of copyright-protected content while giving publishers and content creators control over the presentation of their works. Google compares Australia unfavourably to other countries with more innovation-friendly legal environments, such as the United States, which has 'fair use' protections, and Singapore, which updated its copyright law in 2021 to include exceptions for computational data analysis.

The company warns that the current legal framework in Australia is already impeding its ability to build AI research capacity and investment in the country. Google argues that the legal risks associated with developing AI technologies in Australia could lead to a loss of local talent and investment, resulting in a decline in job creation, investment, and intellectual property.

Google also addresses concerns about the use of copyrighted works to train AI platforms and the ownership of works created using AI systems. It argues that requirements for organisations to explain decisions made by AI should not be overly burdensome. While Google recognises the importance of making AI decisions explainable to build trust and confidence in the technology, it warns that imposing too much transparency would be unworkable and could restrict the social and economic benefits of AI.

Additionally, Google cautions against requiring Australian organizations to only train AI systems on data residing in Australia. The company argues that policies restricting Australian organizations from using offshore computational resources or certain types of data for model training would increase costs for AI innovators in Australia and hinder the tailoring of global AI systems to Australian needs and circumstances.