In December, election workers from across Arizona convened at a Phoenix-area hotel for a groundbreaking training exercise orchestrated by Arizona Secretary of State Adrian Fontes. The session was designed to prepare them for the types of artificial intelligence (AI) attacks they might encounter during the upcoming election cycle in this politically competitive state. Fontes, a Democrat, greeted participants via a video message, emphasizing the importance of the training in honing their skills against new challenges.

Over two days, these workers engaged in complex scenarios generated by AI, simulating various threats from law enforcement operations to technological infiltrations. The exercise plunged them into deepfake scenarios, where they had to navigate through fake crises and misinformation campaigns.

The twist came at the end when Fontes revealed that the person in the welcome video was not him but an AI-generated fake, constructed from publicly available images. This revelation underscored the sophisticated nature of the threats they were learning to combat.

Local Perspectives and AI Challenges

Dana Lewis, the county recorder in conservative Pinal County, found the simulations both enlightening and unsettling. By the end of the training, the sentiment among the workers was to "trust no one," a stark departure from the usual community-oriented approach of election administration.

The exercise highlighted the emerging challenges posed by AI-generated deepfakes, which FBI Director Christopher A. Wray pointed out at a recent national security conference. Wray noted that this election cycle's adversaries, armed with advanced AI technologies, could execute influence campaigns much faster than before.

Training and Preparedness

The exercise was part of a larger initiative by the Brennan Center and the Elections Group, who have been at the forefront of studying AI's impact on elections. They helped develop the training scenarios after many election officials expressed concerns about the lack of direction in handling AI threats. A training document titled, “How Election Officials Can Identify, Prepare for, and Respond to AI Threats,” was created post-event to aid election workers nationwide, although the organizers acknowledged the impossibility of reaching all officials across the country before the elections.

Practical Applications and Real-World Strategies

The exercise also incorporated practical elements, such as assigning teams fictitious counties and limited budgets to mirror real-life constraints. Participants had to make strategic decisions on what security measures to invest in, reminiscent of role-playing games like Dungeons and Dragons.

Following the exercise, Dana Lewis took proactive measures to protect her digital identity, limiting her social media presence and editing her photos to prevent misuse.

Impact and Future Preparations

By the end of the training, participants were equipped with strategies to combat misinformation and verify the authenticity of communications. They learned the importance of multifactor authentication, using secure .gov web domains, and maintaining strong local media relationships to counteract fake news.

Adrian Fontes stressed the challenges of controlling or even accessing many digital platforms where AI-generated fakes might circulate, highlighting the ongoing struggle against these evolving threats.

Conclusion

The Arizona training session marked a significant step in preparing election officials to face AI threats. It combined practical security measures with an understanding of the psychological impacts of AI misinformation, offering a comprehensive approach to safeguarding the integrity of elections. As AI technology continues to advance, such training will become increasingly crucial in ensuring that election officials can protect the democratic process.