Prompt Injection Vulnerability of Consensus Generating Applications in Digital Democracy

Large Language Models LLMs are gaining traction as a method to generate consensus statements and aggregate preferences in digital democracy experiments. Yet, LLMs may introduce critical vulnerabilities in these systems. Here, we explore the impact of prompt-injection attacks targeting consensus...

How Maria Barsallo Lynch Helps Combat the Spread of Misinformation and Disinformation Ahead of the Election

In our most recent episode of Security Nation, we spoke with Maria Barsallo Lynch, Executive Director of the Defending Digital Democracy Project D3P at the Belfer Center for Science and International Affairs at the Harvard Kennedy School, about her work informing election officials of the rise of...

Beers with Talos Ep. #53: Shiny happy election security (and ninjas)

Beers with Talos BWT Podcast Ep. 53 is now available. Download this episode and subscribe to Beers with Talos: If iTunes and Google Play aren't your thing, click here. Recorded May 10, 2019 — Election security has been a dominant headline for some time, so it’s high time we take a look at what th...