Documents released under the Freedom Of Information Act have revealed that the Department of the Prime Minister and Cabinet (PM&C) conducted tests using AI program ChatGPT as far back as early last year.
Documents dated to February 2023 confirmed that the “PM&C has undertaken some small scale AI testing in the past”. Despite this, the documents said a cyber security review of ChatGPT had not been undertaken at the time.
It said the tests included using a chatbot to provide information about the PM&C Enterprise Agreement, an employment contract for PM&C employees. The testing was “undertaken as a proof of concept and is no longer active”.
Although the department hasn’t utilised the program yet, it would consider doing so if a business requirement existed. It also said there was potential for future use in data analytics, particularly for categorising and sorting unstructured datasets.
However, such use had not yet been approved.
“As is the case with the potential procurement and implementation of any new technology, PM&C would need to consider skilled resources, funding and the suitability of any AI program or service used to ensure it meets privacy, security and value-for-money criteria,” the documents said.
Since then, Prime Minister Anthony Albanese has announced that the federal government will further trial AI in partnership with Microsoft. The announcement, which did not mention these earlier tests, said the trials would commence this month.
Experts have raised serious ethical concerns about the trial, particularly regarding potential bias. The government published its interim response to community feedback on its Safe and Responsible AI in Australia discussion paper.
Part of the response included an AI expert advisory body.
Whereas experts have welcomed the paper as a first step, they say further consideration of the ethics of AI systems, alongside the development of codes of conduct for professional practice, are needed.
