How to navigate AI manipulation this election season

An attacker doesn’t have to change a single vote to damage confidence in an election. Sometimes, creating confusion is enough. We can protect the systems that run elections and still leave voters exposed to manipulation.

For example, in 2024, Iranian cyber actors stole private campaign material and attempted to distribute it to influence the election. U.S. intelligence agencies also attributed a fabricated video showing ballots being destroyed in Pennsylvania to Russian actors. These different tactics shared the goal of exploiting division and undermining confidence.

Artificial intelligence makes that challenge more complicated. It can make deception cheaper, faster and easier to personalize, but the risk goes beyond creating more convincing fakes. It can also make it easier for dishonest actors to cast doubt on real evidence by claiming it was generated by AI.

Social engineering exploits how people think, whom they trust and what prompts them to act. During an election, attackers have plenty to work with, including strong emotions, familiar public figures and a constant stream of consequential information. 

Consider a hypothetical call that sounds like a candidate asking for an urgent donation. Or a text claiming your polling location has changed. Or a video appearing to show an election official admitting misconduct. Each exploits a familiar identity or institution, and AI can help attackers produce convincing variations at scale. 

The same approach can target campaign staff. An attacker impersonating a colleague could use a cloned voice to request an authentication code. One successful deception could expose private communications that are........

© The Hill