United States - Massachusetts - Generative AI Guidelines

Overview

The City of Boston's Interim Guidelines for Using Generative AI, version 1.1, were prepared by Santiago Garces, Chief Information Officer for the City of Boston, and published on May 18, 2023 (City of Boston). These guidelines serve as an interim resource for employees of the City of Boston, applying to all City agencies and departments with the exception of Boston Public Schools (City of Boston). The primary purpose is to provide guidance for employees on the responsible use of generative artificial intelligence (AI) tools, acknowledging both their potential usefulness and inherent risks (City of Boston). The guidelines stress that generative AI is a tool, and users remain responsible for the outcomes, emphasizing that 'Technology enables our work, it does not excuse our judgment nor our accountability' (City of Boston). The City encourages responsible experimentation to understand the technology's potential while actively managing its dangers (City of Boston). These interim guidelines are intended to be replaced in the future with more comprehensive policies and standards as the technology evolves and understanding improves (City of Boston).

The guidelines were created in the spirit of helping the city workforce better understand the risks associated with generative AI, encourage responsible experimentation, and support teams interested in learning more about these technologies (City of Boston). The City of Boston was among the first major cities in the United States to release such guidelines, which include principles, sample use cases, and risks to be aware of (City of Boston). The overarching goal is to empower city workers to deliver better, safer, more efficient, and equitable services and products to residents (City of Boston).

Definitions

The guidelines define Generative AI as a set of relatively new technologies that leverage large volumes of data along with machine learning (ML) techniques to produce content based on user inputs, known as prompts (City of Boston). This new content can be written (e.g., ChatGPT or Bard) or visual (e.g., Dall-E) (City of Boston). It is clarified that these tools are not 'actual intelligence in the human sense' but rather 'very sophisticated models that predict what the language, text, or video that satisfies the prompt should be' (City of Boston). Other key concepts, while not formally defined in a separate section, are implicitly addressed through the guidelines, such as 'confidential information' which includes personally identifying information, faces, voices, identifications, and license plates (City of Boston), and the concept of 'bias' in AI outputs that could harm vulnerable populations (City of Boston).

Governance and Institutional Framework

The guidelines were prepared by Santiago Garces, Chief Information Officer, City of Boston (City of Boston). The Department of Innovation and Technology (DoIT) plays a central role in supporting the implementation of these guidelines. DoIT is responsible for staying informed about new technologies and their impact on the city workforce (City of Boston). The guidelines state that DoIT will support events and workshops to help employees learn more about generative AI (City of Boston). Employees are encouraged to contact DoIT for further information on generative AI or if they have applications requiring sensitive information to be used with generative AI (City of Boston). The Mayor's Office of Arts and Culture and the Mayor's Office of New Urban Mechanics are also mentioned as contacts for discussions on the impact of generative AI on arts and society (City of Boston). This distributed contact approach suggests a collaborative governance model for understanding and addressing the broader implications of AI within the city.

Key Focus Areas

• Responsible Experimentation: The guidelines actively encourage City employees to 'try these tools for yourselves to understand their potential' within a framework of responsible experimentation (City of Boston). This approach aims to balance the exploration of new efficiencies and civic engagement opportunities with a clear understanding and control of the tools (City of Boston).
• Data Protection and Privacy: A critical directive is to 'Do not include confidential information in the prompt' (City of Boston). This includes personally identifying information about residents or public servants, faces, voices, identifications, and license plates, especially without consent (City of Boston). The concern stems from the fact that data used in generative AI might be utilized by the companies powering these systems, potentially leading to inadvertent sharing (City of Boston). Employees are advised that if they wouldn't share information in public, they should avoid sharing it in a prompt (City of Boston).
• Accuracy and Fact-Checking: Employees are explicitly instructed to 'fact check and review all content generated by AI, especially if it will be used in public communication or decision making' (City of Boston). The guidelines caution that generative AI can produce inaccurate, outdated, or fabricated information, and it is the user's responsibility to verify accuracy through independent research (City of Boston).
• Transparency and Disclosure: When content is produced with the aid of generative AI, employees must 'disclose when content has been created with AI' (City of Boston). This includes indicating which model and version were used and who edited it if it is used in public communication or decision-making (City of Boston).
• Accountability: The foundational principle is that 'We are responsible for the outcomes of our tools' (City of Boston). Technology does not excuse judgment or accountability (City of Boston). Public sector professionals are trusted to use their judgment to maximize benefits while avoiding negative impacts for the City and its constituents (City of Boston).
• Equity and Inclusion: The guidelines explicitly state the importance of ensuring that AI outputs 'will not be offensive or harmful towards people, particularly vulnerable residents that are susceptible to harm including ethnic and racial groups, diverse gender individuals, and others' (City of Boston). The use and development of AI should support work that addresses damage done to racial and ethnic minorities, people of all genders and sexual orientations, people of all ages, people with disabilities, and other marginalized groups, aiming to uplift these communities (City of Boston).
• Use Cases for Productivity and Communication: The guidelines provide examples of beneficial use cases, such as drafting memos, job descriptions, summarizing text, repurposing content for different audiences, and translation (City of Boston). They also highlight how generative AI can assist in creating visual, audio, and video communications, and help in writing clearer and simpler language by specifying reading levels or audiences (City of Boston). For code generation, caution is advised, requiring users to understand what the code is doing and be familiar with vulnerabilities before using it in production (City of Boston).

Implementation Framework

The Boston Interim Guidelines for Using Generative AI are presented as an 'interim resource' (City of Boston). This implies a temporary status, with the expectation that they 'should be replaced in the future with policies and standards' (City of Boston). The Department of Innovation and Technology (DoIT) is tasked with supporting ongoing learning and engagement by facilitating events and workshops for employees interested in these technologies (City of Boston). The City also encourages employees to share their experiences, thoughts, and concerns via an online form, fostering a collaborative approach to understanding the practical implications and evolving needs related to generative AI (City of Boston). The City of Boston is continuously researching, testing, and implementing infrastructure for the responsible public use of generative AI, with plans to release updated guidelines informed by this ongoing research (City of Boston).

Monitoring and Evaluation

While the guidelines do not lay out a formal, explicit monitoring and evaluation framework with specific metrics or reporting cycles, the City of Boston demonstrates an ongoing commitment to understanding and adapting its approach to generative AI. The guidelines themselves encourage employees to share their experiences, thoughts, and concerns via an online form (City of Boston). This feedback mechanism serves as an informal method of gathering insights into the practical application and challenges of the guidelines. Furthermore, the City of Boston's Digital Service Team is actively researching, testing, and implementing infrastructure for the responsible public use of generative AI (City of Boston). This includes conducting surveys among city employees regarding their use of generative AI tools, their familiarity with the technology, and their training preferences, as evidenced by a survey conducted in spring 2025 (City of Boston). The insights gained from such research and employee feedback are intended to inform and shape future updates to the generative AI guidelines, indicating an iterative and adaptive approach to governance rather than a rigid monitoring system (City of Boston).

Penalties, Liability, and Appeals

The Boston Interim Guidelines for Using Generative AI do not explicitly detail criminal penalties, civil penalties, fines, or an appeals process. Instead, they strongly emphasize user responsibility and accountability. The guidelines explicitly state, 'Generative AI is a tool. We are responsible for the outcomes of our tools' (City of Boston). It further clarifies that 'Technology enables our work, it does not excuse our judgment nor our accountability' (City of Boston). This means that city employees are ultimately accountable for any content produced or decisions made with the assistance of AI, regardless of potential errors or biases generated by the AI system (City of Boston). The implicit understanding is that existing city policies regarding professional conduct, data handling, public communication, and ethical behavior would apply to situations involving generative AI, with the guidelines serving to clarify expectations within this new technological context. The absence of specific penalties within these interim guidelines suggests that any repercussions for misuse or negligence would fall under broader disciplinary frameworks already in place for city employees, rather than unique AI-specific sanctions.

Relationship to Other Instruments

The Boston Interim Guidelines for Using Generative AI are presented as an 'interim resource' that 'should be replaced in the future with policies and standards' (City of Boston). This indicates that they are a precursor to more formal and potentially legally binding instruments. As municipal guidelines, they operate within the broader legal framework of the United States, including federal and state laws, though they do not explicitly reference specific higher-level legislation. However, the principles embedded within the guidelines, such as data privacy and non-discrimination, implicitly align with existing laws and ethical considerations at federal and state levels concerning data protection (e.g., privacy laws) and civil rights (e.g., anti-discrimination laws) (City of Boston). The guidelines also implicitly relate to internal City of Boston policies regarding data security, public records, and employee conduct. They aim to provide specific guidance for a new technology (generative AI) within the existing governmental operational and legal landscape, laying the groundwork for future, more comprehensive regulatory instruments tailored to AI.

International Alignment

The Boston Interim Guidelines for Using Generative AI do not explicitly state their alignment with international standards or other jurisdictions' regulations. However, the underlying principles of responsible AI use, such as transparency, accountability, fairness, and data privacy, are consistent with emerging international discussions and frameworks for AI governance (City of Boston). Many global discussions around AI ethics and regulation, including those from organizations like the OECD (which has cited these guidelines), emphasize similar values to ensure trustworthy AI systems (OECD.AI). While not a direct adoption of international legal texts, the Boston guidelines reflect a global consensus on key ethical guardrails for AI development and deployment, particularly in the public sector. The focus on avoiding harm to vulnerable populations and ensuring equitable outcomes resonates with human rights-based approaches to AI governance seen in various international recommendations (City of Boston).

Implementation Timeline

Compliance Checklist

Sources and References