‘Near certainty’ of AI disinformation next election as ‘significant regulatory gaps’ remain
A Labor-led Senate committee has warned that there is “near certainty” that the upcoming federal election will see attempts to spread false and misleading information with content generated by artificial intelligence.
A Labor-led Senate committee has warned that there is “near certainty” that the upcoming federal election will see attempts to spread false and misleading information with content generated by artificial intelligence.
It has recommended implementing a voluntary scheme of “watermarking” and labelling AI-generated content.
The Labor and Coalition senators were most divided on how far legislation should restrict the production or sharing of AI-generated political material. Labor’s Tony Sheldon, chair of the select committee on adopting AI, recommended in the interim report that such rules be designed to complement ongoing work in the space, including Labor’s misinformation bill and mooted truth in political advertising reforms. Coalition senators James McGrath and Linda Reynolds vehemently opposed the suggestion, saying “distinguishing between truth, opinion, and falseness is an inherently subjective process, and one … appropriately left to the Australian public”.
AI’s place in democracy has been an issue of growing concern as critical developments in generative AI have lowered the skill, time, and resources required for the creation of convincing deepfake audio, images, and videos.
Furthermore, large language models such as OpenAI’s GPT models – which power ChatGPT – could allow for more individually targeted messaging.
This has already played out in elections all over the world: in the US, thousands of New Hampshire presidential primary voters received a call from a fake Joe Biden urging them to skip voting. Before a critical election in Slovakia last year, deepfake audio emerged of a political leader and journalist appearing to discuss illicit campaign strategies.
“Evidence of the use of AI-generated content in the context of a number of overseas elections in 2024 suggested there is a near certainty that the upcoming federal election in Australia will be subject to similar attempts at spreading disinformation,” Senator Sheldon’s report read.
“The committee notes that as acknowledged by the government and numerous submitters to the inquiry, there are significant regulatory gaps in Australia’s capacity to respond effectively to the use of AI in the context of electoral processes.”
The committee called on the government to implement voluntary codes about watermarking and credentialling of AI-generated content before the next federal election. The report noted that this technology was only in its infancy and not foolproof nor tamper-proof.
The Coalition senators were open to the suggestion “in principle” but reserved judgment until the experience of the upcoming US election could provide guidance on the issue. The Greens proposed a “temporary and targeted ban” on political deepfakes in the next election, and independent senator David Pocock suggested a ban of AI-generated material on election material for the next election and subsequently reviewed to develop a “mandatory code” for following elections.
Senator Sheldon also recommended a “thorough review” of potential responses to AI-generated political deepfakes to take place before the election, as well as examination of “mechanisms, including education initiatives, to improve AI literacy”.
More Coverage
The verdict: How the Miles v Crisafulli debate played out
The unanswered question from today’s Queensland leaders’ debate was this: how long can David Crisafulli hold his nerve?
It’s time for a change, ‘evasive’ Crisafulli wins
David Crisafulli was evasive on policy specifics, but the LNP leader has won the second Queensland leaders’ debate solely on his argument that it’s time for a change after a decade of Labor in power.