Four steps to protect our elections from AI interference

I recently co-signed a letter with a host of leading third-sector organisations, as well as consumer champion Martin Lewis and Wikipedia Founder Jimmy Wales, calling on political parties to agree a set of guidelines and guardrails for how they use AI in election campaigns.

AI has developed at a rapid pace in recent years (see the AI generated image on this article) and we are already seeing its impact on election campaigns around the globe, with it enabling the creation of artificial video and audio clips that are hard to distinguish from reality.

This year’s general election will likely be the first where we see the emerging impact AI will have on our politics. That is why the ERS is backing the call, led by Demos, to put protections in place before voters head to the polls.

Democracy can only function if it is grounded in honest, good-faith debate and voters can trust what they see and hear from our politicians.

Open Letter

In a year where globally more voters will head to the polls than ever before, democracy is facing a new and largely untested threat – that of AI-generated content. While there are significant opportunities to harness AI technologies for the benefit of humanity, including for democracy, they also pose major risks to our electoral processes, particularly in voter access to accurate information. The ease with which deepfakes can now be produced could further undermine trust in our democratic processes. Political parties are by no means powerless on this issue and can play a role in protecting election integrity in the run up to the UK election. Therefore, we are calling on all UK political parties to make an agreement committing to four steps which can help protect the UK election:

1. Not using generative Al tools to produce materially misleading audio or visual content that might convince voters into believing something is true when it is not
2. Clearly labelling where generative Al is used to produce audio or visual content in a non-trivial way*
3. Not amplifying materially misleading Al-generated content and being aware that ‘calling-out’ bad actors can sometimes have this impact
4. Ensuring that party staff, members, volunteers and supporters are given clear guidelines for the use of generative Al in election campaigning

We believe UK political parties can show global leadership on this issue, helping to establish good norms around the use of emerging technologies and ultimately protecting democracy in this critical year. If they do not, they risk further diminishing public trust in our political process.

Signed

• Polly Curtis, Chief Executive, Demos
• Sam Gregory, Executive Director, WITNESS
• Jeffrey Howard, Associate Professor and Director, UCL Digital Speech Lab
• Darren Hughes, Chief Executive, Electoral Reform Society
• Martin Lewis, Founder and Chair of Money Saving Expert and the Money and Mental Policy Institute (MMHPI)
• Chris Morris, Chief Executive, Full Fact
• Gina Neff, Executive Director, Minderoo Centre for Technology and Democracy at Cambridge University
• Rashik Parmer MBE FBCS, Group CEO, BCS, The Chartered Institute for IT
• Jimmy Wales, Founder, Wikipedia

* Trivial altering of content is content that is altered or generated in such a way that is inconsequential to the viewer’s perception of it. This is exempt from disclosure under this commitment. This may include edits that do not materially change the implied context or content of an event.

We can’t leave political transparency to the whim of big tech firms: it’s time for a campaign overhaul.