Project summary

I am developing a practical framework that adapts traditional election observation methods to test how AI systems perform during elections. With more voters using LLMs and AI-mediated search as primary information sources, we need a reliable way to monitor the risks. This project gives civil society and policymakers the evidence they need to develop policy recommendations and advocate for safe AI and reduce the risk that AI failures erode democratic legitimacy.

Over the next four months, I will build and pilot an open-access toolkit for domestic election monitors and democracy practitioners. The framework will include prompt libraries covering common election-related information requests, testing protocols, documentation templates, implementation guidance, and recommendations for interpreting and communicating findings. Because elections occur across diverse political and linguistic environments, the project is designed to produce an evaluation framework that can be adapted across contexts.

To build this, I’m combining classic election-monitoring principles with modern AI evaluation practices. I'll be refining the methodology alongside election experts, relationships I've built through a decade of work in elections, including representing the US government as an international election observer in countries like Azerbaijan and Kyrgyzstan. I will also consult experts in AI evaluations and model testing to ensure the methodology is statistically sound and avoids overstating conclusions.

Theory of Impact

Managing existential risk requires high-functioning states that can cooperate, pass smart regulations, and enforce guardrails. If AI systems break the public's trust in elections, then state capacity disappears.

This project tackles x-risk in three ways:

1. If AI-driven misinformation breaks society's shared reality, political systems break down, and globally we will not be able to coordinate on existential threats. Keeping the information ecosystem stable keeps the door open for safety governance.

2. This framework brings safety testing into the wild, across diverse languages and political environments. It creates a massive, independent feedback loop that forces developers to fix structural vulnerabilities before their models become more powerful and autonomous.

3. By training domestic monitors, civil society groups, and ordinary citizens to rigorously test these systems now, we are building the global auditing infrastructure we will desperately need as AI capabilities scale toward critical thresholds.

What are this project's goals? How will you achieve them?

The immediate project output is a practitioner-ready framework that can be used by nontechnical experts to evaluate model performance across election contexts globally. To ensure that it is then implemented and utilized, the project will also include a guided pilot with domestic election observers in an upcoming election - prospectively the October elections in Latvia a country with high digital adoption, a multilingual information ecosystem, and significant geopolitical vulnerabilities. To support adoption and impact beyond this pilot, the tool will be disseminated through the Global Network of Domestic Election Monitors and include guidance materials and training focused not only on how to use the tool to generate evidence, but how to use that evidence to advocate for AI safety in the user's local context.

The ultimate goal is to make AI evaluation a routine component of election observation worldwide, that enables democratic institutions to identify emerging risks early, generate evidence from underrepresented electoral contexts, and drive continuous improvements in the safety of AI systems during elections. More broadly, the project seeks to strengthen the democratic accountability and governance infrastructure needed to ensure increasingly capable AI systems remain subject to independent oversight, a critical prerequisite for managing the broader societal risks posed by advanced AI, making this work an investment in long-term AI governance capacity.

How will this funding be used?

$35,000 allows the project to move beyond a draft framework into a tested, practitioner-ready resource. Funding would support my time developing, coordinating, and refining the framework over approximately four months; expert feedback from AI evaluation and election observation practitioners; structured pilot testing with intended users; translation and localization of materials for at least one non-English election context; and dissemination through practitioner networks.

Who is on your team? What's your track record on similar projects?

The project is led by Genevieve Shea, an AI safety and democracy expert, with more than a decade of experience supporting governments, civil society, and international organizations to respond to emerging governance challenges. She previously directed a $6.5M regional initiative to strengthen citizens' political engagement and government performance in Central Europe. Her expertise in elections has allowed her to represent the US government as an international election observer in several countries, including Azerbaijan, Kyrgyzstan, and Kazakhstan. She is currently an AI Policy Fellow with Successif, where this project was first conceived, and benefits from expert feedback from practitioners in the AI safety community.

What are the most likely causes and outcomes if this project fails?

The risk the project fails to achieve its output goals is very low. The idea was originally conceived through the Successif AI Policy Fellowship and therefore benefits from an accountability structure, including regular check-ins and guidance. The greater risk is adoption. To address this, the project will support pilot testing and iterative refinement, so that it is both practical and useful upon widespread launch. If the project nevertheless fails to achieve meaningful uptake, policymakers, AI developers, election authorities, and AI safety researchers will continue to have limited empirical evidence about how these AI systems perform in real-world elections, particularly small language and under-resourced democracies, delaying the identification of systematic failures and reducing opportunities to improve them before they affect democratic processes and cause serious harm.

How much money have you raised in the last 12 months, and from where?

None