Andreas Cichowicz

THE FUTURE OF AI AND ELECTION INTERFERENCE Roughly half the global population will go to the ballot box this year, and most of those voters will be bombarded by misinformation — including an explosion of increasingly sophisticated AI-generated deepfakes. Here are five reasons for concern: Deepfakes are getting better, and they don’t even have to be perfect: The AI-generated robocalls in Joe Biden’s voice that urged voters to stay home in New Hampshire didn’t fool every voter, but average voters don’t have time to discern authenticity, especially as those fakes get better and better. The “liar’s dividend”: Politicians and pundits exploit the existence of deepfakes to dismiss real evidence as fake, further eroding trust in authentic media​ Opening voters to foreign influence: The distribution of foreign propaganda and manipulation of voters gets easier and easier. Trust lost twice: Deepfakes undermine public trust in the electoral process, but also the use of AI, potentially generating a backlash against beneficial AI applications. Regulatory and technological countermeasures are weak: Federal laws on deepfakes are limited, and only a few states have tried to plug the gap. Technological innovation, meanwhile, has not come up with effective tools to counter deepfakes. The good news is that policy makers and innovators have started to develop solutions. Concepts such as watermarking and entity disambiguation, including Cambrian’s PIDA-BouncerBot IP, can help ensure data authenticity. Regulations with teeth will help authorities and innovators fight back against deepfake distributors, as well. In the meantime, the best answer might just be taking some of the tools we already have – reverse image searches, human fact-checkers, and the ethical compass of most developers – and encouraging more voters to consult with them. @AiCambrian Here’s a list of resources to find out more: cfr.org/blog/election-2024-d… brennancenter.org/our-work/r… journalistsresource.org/home… campaignlegal.org/update/how… californialawreview.org/prin… americanbar.org/groups/leade… sumsub.com/newsroom/deepfake…