At a symposium on the impact of AI last week, Sydney University law professor Kim Weatherall noted new challenges created by placing the power of automation into the hands of more and more people, including those looking to use this power for ill. We can’t just focus on addressing harms after the event, she said, but need to look at the design of the system. She noted concerns about the EU’s proposed AI Act, which struggles to account for broader societal risks that AI developers cannot competently assess.
One risk which Open AI, producer of ChatGPT, could surely have assessed before it released its model to the public is the use of personal information to train the model. Italian regulators recently moved against Open AI for failing to comply with its obligations under the EU’s General Data Protection Regulation. They noted strong privacy concerns with the dataset used to train the firm’s GPT models, as well as with the potential misuse of user data obtained through interactions with the ChatGPT interface. ChatGPT was banned until last week, when Open AI provided greater transparency over use of the data, and the ability for users to opt out of data collection.
However, the steps they’ve taken do not appear to address the privacy problems with the training data, nor with the ability of the model to produce inaccurate information about individuals, with the regulator’s statement noting that it is ‘technically impossible … to rectify inaccuracies.’ These are problems for Australia to note as it seeks to reform the Privacy Act, and this week’s announcement of the government’s intention to install a dedicated privacy commissioner to combat data breaches is a welcome step.
In this week’s newsletter, Monica looks at another problem with the ready availability of the new AI tools – their ability to automate the production of news, or news-like, content, sometimes of questionable veracity. Of course, AI is not the only hot media issue. Ayesha looks at Australia’s performance in the latest World Press Freedom Index (hint: still worrying), Sacha comments on the ACCC’s findings about influencer advertising on social media, and Derek examines ACMA’s recent breaches of the Outsiders program’s climate coverage.
Michael Davis, CMT Research Fellow