19 August 2024

Opinion: TPDC calls for candidates to pledge transparent use of AI in the election

We call on political candidates to sign up to a Pledge for Digital Transparency in Campaigning (DigiTIC), committing to:

  • use content credentials on whatever services they are available; and
  • add a visible disclaimer where GenAI is used to deceptively fake or alter the appearance, voice, or actions of political candidates, election officials, and other key stakeholders in a democratic election

17 August 2024

Tell us how AI shapes our election

Here’s one thing we can do – right now – to protect Australia’s democracy before the next federal election.

By Johanna Weaver, Zoe Hawkins

Generative Artificial Intelligence is here and the question is not if but how it will be used during the next federal election campaign. And how will Australians be informed about its use? With our next federal election looming, these questions require urgent attention.

GenAI is already being used in elections around the world; but this is not objectively good or bad. Like other technologies, GenAI can be used in ways that enhance democracy, undermine democracy, or in ways that don’t impact democracy at all.

We are calling for political candidates to sign up to a pledge for digital transparency in campaigning before the next federal election campaign formally kicks off. Here’s why.

Not all GenAI use is harmful

While much of the public debate focuses on the risks of harm, not all uses of GenAI are harmful, or even noteworthy. GenAI is now embedded in commonly used and widely available content creation and communication tools. These technologies are being used by Australian individuals and businesses every day. Naturally, these tools are equally likely to be used by candidates and political campaigns. Consider, for example, how GenAI embedded in graphic design tools like Canva or Photoshop can help isolate headshots for candidate posters. Or how smaller parties can harness GenAI to get their message out to voters at a speed and scale previously only available to established parties with deep pockets. Ditto how GenAI tools, and the ease of video and image editing, can augment the well-established tradition of political satire.

GenAI undoubtably brings risks, but it is also already an everyday tool that can be a force for good in democracy. An outright ban on the use of GenAI in elections – such as that trialed in Korea – would likely disproportionally and negatively impact smaller parties.

Emphasise proof of authenticity

That said, there is no denying the increasing volume and sophistication of synthetic content online that has the potential to mislead or deceive voters. Democratically moderating a GenAI fueled infosphere is a complex problem. In the immediate term, we need to be pragmatic and prioritise interventions that can build trust between now and election day. In our efforts to preserve the integrity of democratic debate, we need to prioritise elevating authentic content instead of removing misleading content. As we have learned with online safety, while take-downs remain necessary, this wack-a-mole approach is not a substitute for a proactive and targeted solution.

Firstly, it requires forming consensus on what content qualifies for removal (particularly challenging in a political debate). Secondly, it is hard to scale as the volume of online content skyrockets. And, thirdly, malicious actors are constantly iterating their strategies to evade take-down measures. Therefore, rather than relying solely on take-downs, we should also focus efforts on elevating authentic content and empowering Australians to seek out trustworthy voices.

In an election context this means we should urge Australians to seek out and rely upon trustworthy sources, such as candidates’ authorised campaign materials or official advice from the Australian Electoral Commission (AEC).

Signal authenticity from the start

In February, 20 industry players signed a Tech Accord to Combat Deceptive Use of AI in 2024 Elections, known as the Munich Accord. Companies committed to trying to detect and appropriately address the distribution of this type of content on their platforms. This important effort is a cat-and-mouse challenge that will require significant investment and sustained efforts. These efforts at the point of distribution should continue with earnest, but we should also intervene all the way up stream: right at the moment of content creation with the addition of “content credentials” that capture verified source and history of media content. These credentials carry forward a tamper-proof record of who created the content, as well as when, and what – if any – modifications have been made to the content.

Providing trustworthy transparency on content provenance right from the moment of content creation means we can teach Australians to look for proof of authenticity rather than proof of fakeness. This is important as detecting synthetic material is only going to get harder every day as the tech gets better and the quantity of content proliferates.

We need to provide people with a way to identify trusted, verified content. Widespread use of these kinds of credentials by candidates would make it simple for voters to identify and rely on official “authentic” campaign materials. And, the good news is, implementation of this approach by candidates does not have to be complicated.

Call for a pledge

To realise the full potential of content credentials, we need the “authentic” sources of official campaign materials to be transparent about when they are using GenAI and enabling the responsible use of GenAI during elections. As Australians approach the ballot box in an increasingly noisy information environment, they deserve to be able to confidently identify official campaign material and know that such official sources will disclose where GenAI has been used to meaningfully alter content. That is why we call on candidates to sign up to a Pledge for Digital Transparency in Campaigning (DigiTIC). This approach is free, requires no legislating, and is easily implementable today. Many popular content creation tools already include the technology to embed content credentials; candidates just need to turn the feature on. And where content is meaningfully created/altered by GenAI in line with the Munich Accord definition, candidates can just add a simple disclaimer.

After inspiring discussions this week with Audrey Tang and Glen Weyl about the success of Taiwan’s ambitious policies to harness AI for the benefit of democracy, it is clear this pledge is a very small step – but a step in the right direction. Hopefully the first of many. Implementation of the DigiTIC pledge will empower Australians to make informed choices during the election. It’s a win for candidates. A win for political parties. A win for voters. And a win for Australian democracy. Our team at the ANU Tech Policy Design Centre stands ready to work with candidates, political parties, industry and the AEC to realise the DigiTIC pledge. We look forward to hearing from candidates ready to take the lead.

To view the Canberra Times article, click here.