The Limitations of Good Policies: Lessons from Puerto Rico’s Implementation of AI Regulations

Puerto Rico's AI policy follows international good practice on paper. Weak institutions, political rivalry and corruption shape how it was made and limit how it is implemented.

Diego Mojica Student author

Article · 8 November 2026 · 9 min read

The Capitol of Puerto Rico in San Juan at dusk, its dome lit
Photo: Amaury Laporte, CC BY 4.0, via Wikimedia Commons

Introduction

This article examines the development of AI regulation in Puerto Rico. Specifically, it argues that weak institutions and corruption have shaped how the archipelago’s AI policy was made and limit how it has been implemented. Much of the attention to AI regulation has focused on technically advanced and higher GDP state contexts such as the United States, the United Kingdom, China, and EU member states. However, the development of AI regulation in states with weak governing capacity and officials who face the pressure of improper influence is equally important and poses its own set of challenges. It is useful for diplomats engaging in diplomacy for AI governance or working in states with weak governance capacity to understand how these dynamics can affect AI regulation and its implementation in practice.

Puerto Rico’s context

Three features of Puerto Rico’s context have been particularly important in shaping its AI regulation and therefore worth describing at the outset First, its status as an overseas territory of the United States typically leads it to adopt federal standards rather than designing its own as a sovereign state. In the case of cybersecurity, these standards come from sources such as the National Institute of Standards and Technology (NIST). Second, Puerto Rico has a track record of corruption that bears directly on how new policies are carried out.1 Finally, its public institutions operate under austerity imposed by an unelected oversight board established in 2017, illustrating how external pressures have also shaped the successful implementation of these policies.

Puerto Rico's AI policy framework

The core of Puerto Rico's AI regulation is the Policy for the Use of Artificial Intelligence in the Government of Puerto Rico (PRITS-POL-0012), a 130-page document published in 2025 by the Puerto Rico Innovation and Technology Service (PRITS). The document provides five main takeaways regarding the use of AI. First, each agency must establish an internal governance mechanism accountable to its Agency Head, while PRITS retains government-wide oversight and must approve the deployment or significant modification of any AI system. Second, agencies must apply the NIST AI Risk Management Framework to classify each system as low, moderate, or high risk based on its potential impact on individual rights, public safety, or access to essential services. Those systems which are deemed high-risk are subject to stricter assessment. Third, the policy mandates fairness, transparency, and human oversight which requires public disclosure of AI use, explainability for high-risk decisions, and a ban on "black-box" systems in legal, financial, or civil contexts. Fourth, all AI procurement must be coordinated with and approved by PRITS, with vendor contracts required to include enforceable security and privacy clauses. Finally, noncompliance can result in fines, contract termination, or disqualification from future contracts under Law 40-2024, and agencies have thirty days from the policy's effective date to submit an initial compliance report to PRITS. Notably, this AI policy was issued under the pre-existing Law 75-2019, which created PRITS, and Law 40-2024, the archipelago's cybersecurity law, rather than under dedicated AI legislation.

These reforms are a positive first step regarding AI governance which, at least on the surface, align with international good practice and with the needs echoed by Puerto Rico's business sector. In a recent article for a Puerto Rican newspaper, journalist Eva Lloréns Vélez argues that the need for AI regulation has taken centre stage. Specifically, she contends that the archipelago needs substantial investment in talent, institutional reorganisation, data, and new regulatory capacities alongside a regulatory environment that "guides rather than limits advancements in AI" for Puerto Rico's economy to properly integrate AI. It is this analysis presented by Vélez that will be examined against the way in which the framework came into being and against recent events in Puerto Rico's government that may undermine the outcome she envisions.

A Governor and a Senate at Odds

Puerto Rico’s AI framework emerged as much from a rivalry within the governing New Progressive Party (PNP by its Spanish Initial) as from institutional design. Since 2025, the legislature, led by Senate President Thomas Rivera Schatz, has advanced a series of bills to build a legal and institutional architecture for AI. Of this series of bills, three stand out. The first is Senate Bill 68. Senate Bill 68, which would create a Government AI Officer and an advisory committee attached to PRITS, was approved by the Puerto Rican Senate on the 7th of April 2025. However, according to the Puerto Rican outlet Platea, this bill is still awaiting a House vote as of the 10th of June 2026. The second bill of interest is House Bill 347. This bill would have made PRITS the regulator of AI use across the executive branch, with power to impose fines of between $2,500 and $10,000. Despite passing both chambers, the bill did not become law due to a pocket veto by Governor Jenniffer González. The third and final bill of interest is Senate Bill 769. This bill would have created an AI Development and Innovation Institute with an annual fund of $10 million, but Governor González also vetoed it expressly on fiscal grounds. Although González vetoed these two laws, she did sign narrower measures which included laws requiring disclosure of AI in electoral communications and guaranteeing citizens the right to know when they interact with automated systems in public services. However, her new measures did not create a dedicated AI regulator or budget. Even at the earliest stage, the then director of PRITS, Antonio Ramos Guardiola, told a Senate hearing in March 2025 that he supported the AI Officer, but he expressed his reservations about a new advisory council.

The outcomes of these bills are difficult to separate from the relationship between Governor González and Rivera Schatz, who are both prominent members of the PNP. In April 2026, González described the Senate president as “the greatest obstacle facing her administration”, and Rivera Schatz responded by conceding that he had indeed obstructed decisions he considered “improper”. This intra-party dispute between Schatz and González has run through nominations and agency management for much of her term, and PRITS has been one of its main arenas. With speculation of Rivera Schatz plausibly launching a primary challenge against González’s gubernatorial re-election in 2028, further obstruction by Schatz is expected.

As a result of this contest, two practical consequences followed. First, with legislation deadlocked, reform could proceed only through administrative policy and existing powers, so the executive was unable to create offices, budgets, or penalties that require legislative approval. Second, the fines in the PRITS policy rest on pre-AI cybersecurity laws that do not cover all of the risks that AI poses and treat AI as a cybersecurity issue. As a result, the responsibility for the whole framework falls on one pre-existing agency, rather than on the new bodies the legislature proposed.

The Impact of Weak Institutions and Corruption

PRITS is poorly placed to carry the burden of implementing AI regulations. In September 2025, the governor herself stated that the agency had not had a permanent director in five years, and she announced plans to return it to the Office of Management and Budget. Between January and September 2025 alone, at least five officials led or were nominated to lead the agency. Those nominated for the position during this period were Ramos Guardiola, Luis Rodríguez Vega, Martín Jiménez Morales, Rubén Quiñones Milán and Poincaré Díaz Peña.

Concerns about PRITS's ability to exercise independent oversight have been further heightened by allegations of political interference in its technology procurement decisions. In April 2025, PRITS director Antonio Ramos Guardiola rejected a proposed contract with XUVO Technologies to operate a digital licensing platform for the Department of Health. XUVO had previously been disqualified from a competitive procurement process for a similar platform, and Guardiola objected to the new contract on those grounds. Three days after his rejection, the governor withdrew the nomination of Guardiola to lead PRITS, and the contract was approved the following day under an interim director. An investigation by the Center for Investigative Journalism (CPI) alleged that Guardiola's opposition to the contract had prompted his removal, although the governor's spokesperson did not confirm any connection.

Due to the frequent turnover of PRITS directors, the agency lacks stable leadership at the moment it is being asked to approve every government AI system and procurement. Its capacity to fulfil this role for AI governance is further undermined by a high vacancy rate of positions across the agency: a report by El Nuevo Día found that roughly 55% of roles remained unfilled as of January 2026.

Beyond these allegations of corruption, the constraints imposed by the unelected Financial Oversight and Management Board (FOMB) add a further layer of difficulty. Established in 2016 following Puerto Rico's debt crisis, the FOMB has overseen austerity measures affecting public services, including education. According to Meléndez Badillo, these measures have caused the closing of around 300 schools within its first two years of operation, while further cutting funding to the University of Puerto Rico. This reduction in spending on educational institutions that would otherwise teach AI literacy further could narrow the pool of talent on which the regulatory capacity Lloréns Vélez calls for depends.

Further Limitations to Implementation Caused by Lack of Transparency

Limited transparency makes it difficult to assess whether Puerto Rico's new AI regulations are being implemented effectively. Although the PRITS policy required agencies to submit initial compliance reports within thirty days, there is currently no publicly available information on these submissions, approvals of AI systems or enforcement actions. It is therefore unclear whether the relevant information has not been collected, has not been published or has been withheld. Recent changes to Puerto Rico's transparency legislation may make it harder to resolve these uncertainties.

In December 2025, Governor González signed legislation amending Puerto Rico's Transparency Law (Law 141-2019). Passed without public hearings and over the objections of a coalition of more than twenty organisations, the bill doubles the time agencies have to answer information requests and requires that all agency officials be notified whenever a document is requested, and removes the option of requesting data in open formats. Furthermore, agencies may declare documents confidential by regulation without justifying the decision and may offer inspection of records at their offices, in some cases for as little as one day. Senate President Thomas Rivera Schatz, who introduced the legislation, defended the changes by arguing that the bill removes no one's rights and seeks a reasonable balance between citizens' requests and agencies' capacity to respond. These restrictions could make it more difficult to scrutinise government agencies' implementation of AI regulations.

The lack of publicly available implementation data prevents a definitive assessment of the effectiveness of PRITS's AI regulations. It remains unclear whether this reflects delays in implementation, limited administrative capacity or inadequate public disclosure. Nevertheless, the evidence presented in this article identifies significant risks to implementation, including political disagreements, unstable leadership, fiscal constraints and allegations of interference in procurement decisions. Together, these weaknesses raise serious questions about whether Puerto Rico possesses the institutional capacity and independence needed to translate its AI regulatory framework into effective oversight.

Reflections on Puerto Rico’s Case

Puerto Rico's experience suggests that adopting international good practice is the easier part of AI governance. If the archipelago is to move from a well-drafted policy to a functioning regime, it will need stable leadership at PRITS, legislative backing, funding for the offices on which the policy relies, and safeguards against political interference in procurement. Future research should track whether agencies file their compliance reports, how PRITS uses its approval powers, and whether the implementation difficulties anticipated in this article materialise, a task that will depend on researchers' ability to access the relevant information. For other jurisdictions seeking to implement successful AI regulation, the case of Puerto Rico suggests that rules must be accompanied by the building of resilient and responsive institutions.

Notes

  1. For more specific information on Puerto Rico’s corruption scandals during the 20th and 21st century, see Chapters 10 - 14 of Meléndez Badillo, Jorell A. Puerto Rico: Historia de Una Nación. ↩

Perspectives are the authors' own views and do not represent their governments or institutions.

References

  1. Burgos Alvarado, Cindy A. "14 medidas legislativas para el uso y regulación de la inteligencia artificial en Puerto Rico." Platea PR, June 10, 2026 (updated continuously). 14 medidas legislativas para el uso y regulación de la inteligencia artificial en Puerto Rico | Platea PR.
  2. Dávila Lozada, Gabriela A. "Senado aprueba proyecto de ley que agrava el acceso a información en Puerto Rico." Centro de Periodismo Investigativo (Puerto Rico), Octubre 14, 2025. Senado aprueba proyecto de ley que agrava el acceso a información en Puerto Rico | Centro de Periodismo Investigativo.
  3. Díaz Tirado, Adriana, and Yaritza Rivera Clemente. "Jenniffer González abre la puerta oficialmente a una primaria contra Thomas Rivera Schatz: 'Nos enfrentaremos.'" El Nuevo Día (Puerto Rico), March 13, 2026. Jenniffer González abre la puerta oficialmente a una primaria contra Thomas Rivera Schatz: “Nos enfrentaremos”.
  4. Generative AI Playbook. U.S. Department of State, 2026. https://www.state.gov/wp-content/uploads/2026/07/DOS-Generative-AI-Playbook_July-2026.pdf.
  5. La Ley de Inteligencia Artificial del Gobierno de Puerto Rico, PS 0068, 20ma. Asamblea Legislativa 1ra. https://aldia.microjuris.com/wp-content/uploads/2025/04/Descarga-el-Entirillado-con-Enmiendas-del-PS68-.pdf
  6. Lloréns Vélez, Eva. "Puerto Rico Frente a La Inteligencia Artificial: Economía, Empleo y El Nuevo Mandato de Política Pública." Metro Puerto Rico (Puerto Rico), July 9, 2026. https://www.metro.pr/negocios/2026/07/09/puerto-rico-frente-a-la-inteligencia-artificial-economia-empleo-y-el-nuevo-mandato-de-politica-publica.
  7. Meléndez Badillo, Jorell A. Puerto Rico: Historia de Una Nación. Primera edición. Translated by Aurora Lauzardo. Editorial Planeta Mexicana, Princeton University Press · Planeta, 2024.
  8. Metro Puerto Rico (Puerto Rico). "Dan paso a crear política pública sobre el uso de inteligencia artificial en el gobierno de Puerto Rico." March 15, 2025. https://www.metro.pr/noticias/2025/03/15/dan-paso-a-crear-politica-publica-sobre-el-uso-de-inteligencia-artificial-en-el-gobierno-de-puerto-rico.
  9. Metro Puerto Rico (Puerto Rico). "Gobernadora envía cuatro nombramientos al Senado." April 30, 2025. https://www.metro.pr/noticias/2025/04/30/gobernadora-envia-cuatro-nombramientos-al-senado/.
  10. Metro Puerto Rico (Puerto Rico). "Jenniffer González propone eliminar PRITS y devolver funciones a la OGP." September 15, 2025. https://www.metro.pr/noticias/2025/09/15/jenniffer-gonzalez-propone-eliminar-prits-y-devolver-funciones-a-la-ogp/.
  11. Metro Puerto Rico (Puerto Rico). "Jenniffer González tilda a Thomas Rivera Schatz como 'el mayor obstáculo' de su administración." April 24, 2026. https://www.metro.pr/noticias/2026/04/24/jenniffer-gonzalez-tilda-a-thomas-rivera-schatz-como-el-mayor-obstaculo-de-su-administracion/.
  12. Metro Puerto Rico (Puerto Rico). "Pica y se extiende la pugna: Rivera Schatz defiende su 'obstrucción' y arremete contra la gobernadora y su equipo de trabajo." April 24, 2026. https://www.metro.pr/noticias/2026/04/24/pica-y-se-extiende-la-pugna-rivera-schatz-defiende-su-obstruccion-y-arremete-contra-la-gobernadora-y-su-equipo-de-trabajo/.
  13. Política Para El Uso de Inteligencia Artificial En El Gobierno de Puerto Rico. PRITS-POL-0012. Puerto Rico Innovation & Technology Service, 2025. https://docs.pr.gov/files/prits/Pol%C3%ADticas/PRITS-POL-0012%20-%20Politica%20para%20el%20uso%20de%20Inteligencia%20Artificial%20en%20el%20Gobierno%20de%20Puerto%20Rico.pdf.
  14. Primera Hora (Puerto Rico). "Rivera Schatz le tira la raya a la gobernadora." Septiembre de 2025. https://www.primerahora.com/noticias/gobierno-politica/notas/rivera-schatz-le-tira-la-raya-a-la-gobernadora/.
  15. Ruiz Kuilan, Gloria. "Falta de personal en PRITS: nuevo director afirma que el 55% de los puestos están vacantes." El Nuevo Día (Puerto Rico), January 13, 2026. https://www.elnuevodia.com/noticias/gobierno/notas/falta-de-personal-en-prits-nuevo-director-afirma-que-el-55-de-los-puestos-estan-vacantes/.
  16. Ruiz Kuilan, Gloria. "Jenniffer González convierte en ley proyecto que restringe acceso a información pública." El Nuevo Día (Puerto Rico), December 14, 2025. https://www.elnuevodia.com/noticias/gobierno/notas/jenniffer-gonzalez-convierte-en-ley-proyecto-que-restringe-acceso-a-informacion-publica/.