This study presented the findings of a pilot survey on artificial intelligence safety, applied to students, professors, and researchers in Yucatán.

The main objective of this pilot study was to evaluate the effectiveness of the selected questions in measuring the key concerns of the study population, serving as preparation for a broader and more detailed national survey on AI safety in Mexico.

Adapted from ESPAI 2023 by AI Impacts and inspired by the article “Thousands of AI Authors on the Future of AI,” we explored ethical, regulatory, and social concerns. The results revealed consensus around the need for government regulation and the creation of a specialized AI Safety agency in Mexico.

This study was carried out in collaboration with Centro Geo and the Universidad Politécnica de Yucatán (UPY). The idea originated from the weekly meetings of the AI-Safety course, proposed and facilitated by Silvia and Ángel Tenorio. During these meetings, we analyzed both the Blue Dot Impact material and other key documents, including the article “Thousands of AI Authors on the Future of AI,” which was essential in adapting globally tested methodologies to the local context.

Key contributions in this process:

Angel Tenorio from AI Safety México Project:

He introduced the existence of “Thousands of AI Authors on the Future of AI,” verified that the selected questions were properly translated to maintain the meaning of the original article, divided the tasks among participants to ensure project goals were met, and held periodic meetings to follow up on task progress.

Silvia Fernandez:

She proposed a pilot experiment based on the paper, covering the reading and analysis of the article, search and translation of the original surveys, selection and adaptation of questions, sample selection, survey administration, and result analysis. The project was successfully carried out by Janeth and Valeria, with Silvia accompanying them throughout the process and reviewing the results at each stage.

The role of UPY and Centro Geo consisted in offering a professional internship period to develop the project: two positions were opened, which were filled by Valeria Ramírez and Janeth Valdivia for two months. In the first stage, we deepened the study by reviewing the data collection method, translating it into Spanish, and selecting — together with Silvia and Ángel — the most relevant questions from ESPAI for the local survey. Valeria designed the questionnaire based on these questions. Additionally, Janeth developed a section to assess perceptions of AI safety with a focus on local socioeconomic aspects.

Being mindful of time, the survey was launched, receiving responses from 36 participants from various research centers, including the Center for Research in Mathematics, the Faculty of Mathematics at UADY, CINVESTAV, and mainly UPY. In the second stage, with the results in hand, individual topics were selected for deeper development. Valeria focused on “Perception of AI Safety in Mexico,” “AI Risks,” and “How Much Should Society Prioritize AI Safety Research Compared to the Current Level?” while Janeth Valdivia developed “Levels of Concern About AI Scenarios in 30 Years,” “Causes of AI Progress,” “AI Risk Perspectives in 20 Years,” and “Perception of AI Safety in Mexico.”

The outcome of this collaborative work was two research reports delivered to both institutions. Although initially we had considered presenting the findings jointly at a conference, limited time prevented us from integrating both projects. Therefore, only Janeth’s work was submitted to the Congress on Artificial Intelligence (COMIA). We deeply recognize and value the quality and rigor of Valeria’s report; ideally, both contributions would have been combined, as her perspective would have significantly enriched the study. Nevertheless, the experience of submitting a paper for international peer review for the first time was highly enriching for the entire team.

The reviewers’ feedback at COMIA highlighted the importance of grounding the analysis in the local context and its contribution to advancing AI regulation initiatives. We also humbly take their suggestions for improvement to scale this project in the future.

Reviewer #5

“This paper is a welcome and much-needed initiative to explore public and academic perceptions of AI risks in Mexico. I applaud your commitment to not only adapt global tools to a local context, but also to make the findings legible to both policymakers and researchers. You also correctly highlight that some AI risks—such as authoritarian misuse and socioeconomic inequality—resonate particularly strongly in the Mexican context. I encourage you to scale up this effort, either by expanding the sample to other regions or by conducting interviews and qualitative follow-ups. You could also consider publishing a short policy brief based on this paper—your findings and the clear public demand for regulation are highly relevant to current legislative debates.”

Reviewer #8

“However, one of the main weaknesses of the paper is the limited sample size: only 36 individuals participated in the survey. While acceptable for a pilot study, this small number restricts the generalizability of the results and limits the strength of any conclusions. Moreover, while descriptive statistics and visualizations are well-presented, the methodology would benefit from a clearer explanation of how survey reliability and validity were assessed. Finally, the paper could be improved by offering more detailed reflections on how these findings might inform national policy or lead to tangible next steps for AI governance in Mexico.”

After the peer review process, the paper was accepted for presentation as a poster and will be published in Research in Computing Science, an open-access scientific journal with international peer review, published by the National Polytechnic Institute.

This pilot study in Yucatán contributes to the academic and policy dialogue on AI safety in Mexico, laying the groundwork for future research and the development of policies that promote the safe and ethical implementation of AI, aligned with both local and international social and cultural values. The significant consensus in the survey regarding the importance of government regulation and support for a specialized AI safety agency in Mexico is crucial, as it highlights the need for an informed governance strategy to manage AI risks while harnessing its potential benefits.

Such progress would not have been possible without the strong institutional collaboration between Centro Geo and Universidad Politécnica de Yucatán, whose joint efforts — with the participation of Dr. Silvia Fernández, Dr. Alejandro Molina-Villegas, and Dr. Óscar Sánchez Sordia — provided the necessary support for the execution and presentation of the project. Furthermore, the media attention received reflects the growing public interest in these issues and highlights the importance of fostering continued academic partnerships to advance AI Safety advocacy and research in Mexico.

Acknowledgments in this project:

Janeth Valdivia:

“This was a team effort that would not have been possible without the commitment and collaboration of everyone involved. I deeply appreciate the valuable support of the external reviewers from Centro Geo, Drs. Silvia Fernandez, Alejandro Molina-Villegas, and Óscar Sánchez Sordia, who contributed their methodological and academic expertise to strengthen the quality of the study. Their critical review and technical guidance were fundamental in validating the instruments and interpreting the results.

I also extend special thanks to Engineer Ángel Tenorio, project advisor from AI Safety México, for his conceptual guidance and commitment; his support significantly contributed to the strength and projection of the work.

I am also grateful to Valeria Ramírez, my UPY colleague, with whom I worked as a team on the foundation of this study, and finally to the COMIA reviewers who encouraged me to continue using my skills in favor of AI Governance.”

Valeria Ramírez:

“I am deeply grateful to everyone involved in this project for their passion, enthusiasm, and, above all, for the constant support provided throughout the development of this work. Their feedback was key to ensuring a quality outcome.

In particular, I extend my thanks to Drs. Alejandro Molina, Óscar Sánchez, and Silvia Fernández, who welcomed us at Centro Geo and shared their methodological expertise.

I also thank Engineer Ángel Tenorio for his valuable guidance and for sharing key articles and resources throughout the process.

Finally, I thank Janeth Valdivia, my partner in the preparation of this work titled ‘AI Safety in Mexico’, with whom I collaborated in a committed and collaborative manner.”

Listen a bit about: AI-Safety México: AI-Safety México: una encuesta piloto en Yucatán

References

AI Impacts. (2023a, noviembre 28). About. https://aiimpacts.org/about/

Thousands of AI authors on the future of AI. (2024). arXiv. https://arxiv.org/html/2401.02843v

AI Impacts. (2023b). 2023 expert survey on progress in AI [AI Impacts Wiki]. https://wiki.aiimpacts.org/ai_timelines/predictions_of_human-level_ai_timelines/ai_timeline_surveys/2023_expert_survey_on_progress_in_ai

Valdivia, J. (2025, mayo). AI safety en México, una encuesta piloto en Yucatán [Podcast]. En AI Safety en LATAM. Notebook LM. https://open.spotify.com/episode/7lun4sk8BwhzO4MGNM5k6V?si=dde34b84c9424b7b