This year is a politically momentous one, with almost half the world voting in elections. Generative AI may revolutionise our political information environments by making them more effective, relevant, and participatory. But it’s also possible that they will become more manipulative, confusing, and dangerous. We’ve already seen AI-generated audio of politicians going viral and chatbots offering incorrect information about elections.

This report, produced in partnership with University College London, explores how synthetic content produced by generative AI poses risks to the core democratic values of truth, equality, and non-violence. It proposes two action plans for what private and public decision-makers should be doing to safeguard democratic integrity immediately and in the long run:

In Action Plan 1, we consider the actions that should be urgently put in place to reduce the acute risks to democratic integrity presented by generative AI tools. This includes reducing the production and dissemination of harmful synthetic content and empowering users so that harmful impacts of synthetic content are reduced in the immediate term.
In Action Plan 2, we set out a longer-term vision for how the fundamental risks to democratic integrity should be addressed. We explore the ways in which generative AI tools can help bolster equality, truth and non-violence, from enabling greater democratic participation to improving how key information institutions operate.