Opinion
Elections and AI: When anything can be faked, even the truth becomes suspect
"For decades, the central question was whether the ballot boxes were secure," writes Microsoft Israel's Government Affairs Director, Noa Givon. "Today, we must ask an additional question: Does the public believe they are?"
At their core, elections are contests for public trust.
Candidates seek to persuade, campaigns seek to influence, and voters must decide whom to believe. Artificial intelligence adds a new question: not only who said something, but whether it was ever said at all.
In a world where convincing videos, images, and voice recordings can be fabricated within minutes, distinguishing reliable information from manipulation is becoming increasingly difficult. What is at stake is not only the integrity of individual elections, but public trust in democratic institutions, which is already steadily eroding.
We have seen how this can play out in elections around the world. In Slovakia, two days before the 2023 election, an AI-generated audio recording began circulating in which one of the leading candidates could supposedly be heard discussing how to rig the vote. In the United States, ahead of the 2024 New Hampshire presidential primary, thousands of voters received robocalls featuring a fabricated version of President Joe Biden’s voice urging them not to vote.
Romania illustrates just how consequential a crisis of trust can become. In December 2024, only days before the runoff, Romania’s Constitutional Court annulled the first round of the presidential election and ordered the process to begin again. The decision was not prompted by ballot-box tampering or a breach of vote-counting systems. According to declassified intelligence documents, a coordinated, inauthentic influence campaign on social media had propelled a fringe candidate to the top of the polls within weeks.
The case set one of the most significant precedents in recent years: without a single vote being altered, doubts about the integrity of the electoral process were enough to bring it to a halt.
These incidents also had one critical feature in common: timing. Each occurred in the final days before voters went to the polls, when there was almost no time to verify the information, refute it, and ensure that the facts reached the public.
Yet the picture is not entirely bleak. Some countries have managed to counter AI-generated falsehoods with relative success. Taiwan and Finland have invested for years in media literacy, critical-thinking education, and rapid verification mechanisms developed in partnership with fact-checking organizations and news outlets. When manipulated content circulated in those countries, its impact was more limited. The lesson is clear: societal resilience must be built in advance, not on election day.
One of the central challenges of elections in the age of AI is not the fake itself, but the shadow it casts over the truth. When videos and recordings can easily be fabricated, people may begin to doubt genuine information as well. At the same time, anyone confronted with an authentic recording of a problematic statement can dismiss it as an AI-generated fake. The existence of evidence no longer guarantees that it will be believed.
Speed is equally important. False content can reach millions of people within minutes, while checking the facts, verifying the source, and issuing an official response all take time. In the final days of an election campaign, the gap between publication and refutation can prove decisive. An accurate correction that arrives too late may not undo an impression that has already taken hold.
No single actor can meet this challenge alone. Technology companies are developing tools to increase transparency and help verify the origin and authenticity of digital content. They are also establishing shared principles and practices for responding to deceptive AI-generated material. But technology is not a silver bullet. Effective preparation requires cooperation among election authorities, government agencies, technology companies, political parties, and the media, as well as the ability to respond quickly when suspicious content emerges.
Cybersecurity must also be part of this effort. Political campaigns and election infrastructure need protection against phishing, data leaks, and targeted cyberattacks that can compromise sensitive information or fuel further manipulation.
Our information-consumption habits must evolve as well. Asking whether a video “looks real” is no longer enough. We must ask who published it, whether it appears on an official channel, whether credible sources corroborate it, and whether it is being presented in its full context. Digital literacy and critical thinking are now essential elements of a society’s ability to withstand influence operations.
Ahead of Israel’s elections, we will not be able to prevent every piece of fabricated content or every attempt to manipulate the public debate. We can, however, prepare now to limit their impact. This should include rapid official verification channels operated by the Central Elections Committee, agreed response protocols among election authorities, political parties, technology companies, and the media, and the adoption of transparency tools that allow the public to examine where digital content originated.
For decades, the central question was whether the ballot boxes were secure. Today, we must ask an additional question: does the public believe they are?
Noa Givon is Government Affairs Director at Microsoft Israel.
