Generative artificial intelligence is a powerful tool. Not only can chats be conducted convincing us that we are talking to a real person. Images, videos or human voices can also be generated and scenarios conveyed that do not correspond to the facts.

The ITA has examined risks and opportunities of generative AI on behalf of the Austrian Parliament. The final report is now available on the parliament's website, the Executive Summary is also available in English. ‘Our goal was to examine how generative AI can be used by democratic systems. However, this also requires an awareness that AI cybercrime and deepfakes pose a serious threat to democracy,’ emphasises ITA Director Michael Nentwich.

Seizing the right opportunities

The final report, which is now available, covers topics such as digital sovereignty, hybrid threats and AI cybercrime, as well as possible options for dealing with AI. The authors propose a variety of accompanying measures to mitigate risks and make the new technology compatible with democracy.

Specifically, they propose collecting examples of the successful use of generative AI in politics and making them available on a kind of ‘democracy platform’. This platform could also play an important role in the context of initiatives on AI literacy. ‘Above all, it is important to take people along,’ says Michael Nentwich. ‘We as a society should come to an understanding about how democratic debates will be conducted in the future and what role AI applications should play in them.’ Fact-checking and labelling of AI content are urgently needed to minimise misinformation. Other measures proposed include the establishment of a parliamentary commission of inquiry on ‘Democracy and AI’, the development of a code of conduct for AI in politics, and nationwide citizens' forums on fundamental issues of democracy.

Potential for damage significant

Generative AI can contribute to the mass dissemination of disinformation. Microtargeting thus not only offers opportunities but also poses dangers for democratic systems, because it can be used to secrently influence electoral decisions. To avoid this, the study authors propose a general ban on political microtargeting of voters using AI. Environmental damage and consequences for the climate, due to the extremely high consumption of water, energy and land, as well as unethical working conditions in the training and production of AI, are also major problems.

New challenges for Austrian politics

The ITA researchers conclude that parliament, as a central actor in democratic politics, must address these controversial issues. ‘To ensure the best possible functioning of the political system in Austria and the EU, measures must be discussed on a broad basis,’ both inside and outside parliament’, says Nentwich. The worldwide interconnectedness of tech companies, most of which are based outside Europe, and the relative slowness of regulation compared to technological development make this a crucial challenge that can only be overcome through joint European cooperation.