AI Use Policy: A Template for Congressional Offices

Written By Aubrey Wilson

Download (PDF)

Introduction

This template is a starting point and should be customized for each office's specific needs, institutional requirements, and operational context. Consult with House Information Resources or Senate Sergeant at Arms for the latest institutional guidance before implementation.

[Highlighted] sections indicate what needs to be customized by an office that adopts this template.

[Office Name] AI Use Policy

Last Updated: [Date]

Contents

  • Purpose and Scope

  • Policy Philosophy

  • Prohibited Uses

  • Best Practices

  • Compliance

  • Knowledge Sharing

  • Training and Support

  • Acknowledgement Agreement

  • Appendix: Recommended Use Cases

Purpose & Scope

This policy establishes guidelines for the responsible use of artificial intelligence (AI) tools in [Office Name] to enhance productivity, improve constituent services, and support legislative work while maintaining security, privacy, and ethical standards. This policy supplements and must be read in conjunction with guidance from [House Information Resources/Senate Sergeant at Arms] and all applicable institutional policies.

Policy Philosophy

Our office embraces innovation and recognizes AI as a powerful tool to enhance our work serving constituents and advancing legislative priorities. This policy establishes clear boundaries and best practices to ensure responsible use while encouraging exploration of beneficial applications.

Prohibited Uses

AI MAY NOT be used for:

  • Replacing human judgment on:

    • Policy positions or voting decisions

    • Constituent service case resolutions

    • Hiring or personnel decisions

    • Final editorial decisions on public-facing content

  • Processing personally identifiable information (PII), including but not limited to:

    • Social Security numbers, addresses, phone numbers, or email addresses

    • Passport numbers, driver’s license numbers, or any other form of ID

    • Constituent case details or personal information

    • Financial information, account numbers, or tax records

    • Health information or medical records

    • Immigration status or documentation details

  • Handling classified or sensitive information including:

    • All classified materials

    • Whistleblower information

    • Security protocols or emergency procedures

    • Passwords, backup recovery codes, or other cyber credentials

    • Staff personnel information

  • Creating misleading content such as:

    • Deepfakes or manipulated media

    • False statements attributed to the Member or other officials

    • Deceptive social media content

    • Fabricated statistics or sources

NEVER use free or personal AI accounts to supplement official business. Personal AI accounts sit outside the office's security controls, data-handling agreements, and records policies. Only AI licenses, paid for by the office and with security settings enabled, are allowed for use within the office.

NEVER use foreign-based AI models, including models developed in China such as DeepSeek, Qwen, and Kimi. Foreign-based services may store your prompts and files on servers outside U.S. jurisdiction, where they are subject to foreign laws and surveillance.

ALWAYS consult with your office’s IT director before working with AI coding agents, such as Claude Code and OpenAI Codex, or providing your AI model with access to other workplace tools and apps.

ALWAYS opt out of AI model training on your data. Enterprise services such as Microsoft Copilot, Google Gemini (through Google Workspace), and ChatGPT Enterprise are opted out by default, so no action is needed other than to sign in with your official work account. Personal-tier subscriptions, even those approved by the chamber, such as ChatGPT Plus/Pro and Claude Pro, may use your conversations for training by default, so every user must check the setting on their own account. To enable privacy settings on the following platforms:

ChatGPT Plus/Pro:

  • Click your profile icon and open Settings.

  • Go to Data Controls.

  • Turn off "Improve the model for everyone."

Claude Pro:

  • Click your profile icon and open Settings, then Privacy.

  • Turn off "Help improve Claude."

Privacy settings are not retroactive. Data already used for training cannot be removed.

Best Practices

Before using AI:

  • Verify the tool is approved by institutional IT guidance set by [House Information Resources / Senate Sergeant at Arms]

  • Assess the sensitivity of information you plan to input and cross reference the use case with the above restrictions outlined in this document as well as [House / Senate] policy.

  • Consider the purpose and ensure it aligns with office goals

During use:

  • Never input sensitive information - when in doubt, don’t include it

  • Avoid for mathematical calculations and use a calculator instead

  • Beware of sycophancy - models will always agree with you by default

  • Use general examples instead of specific constituent cases or political scenarios

  • Iterate and refine prompts for better results

After use:

  • Always fact-check outputs - AI can produce inaccurate references, incorrect citations, and completely falsified information

  • Edit and personalize to ensure AI-generated content reflects office voice

Compliance

  • Individuals are responsible for the content they produce, use, and sign-off on

  • Staff must report suspected misuse of AI tools

Knowledge Sharing

Staff are encouraged to participate in office-wide information sharing of AI use cases and lessons learned. Such insights are vital contributions to the team’s overall awareness and understanding of these tools’ applicability in the Congressional workplace.

Training & Support

All staff are strongly encouraged to pursue training on the use of AI tools.

 

Acknowledgment: I have read, understood, and agree to comply with this Generative AI Use Policy.

Staff Member: _____________________________ Date: _________

Signature: ________________________________

 

Appendix

Recommended Use Cases

While not exhaustive, beneficial applications of institutionally approved AI tools may include:

  • Research and Analysis:

    • Background research on policy topics

    • Summarizing public documents and reports

    • Analyzing trends in public data

    • Finding and aggregating online information sources

    • Exploring voter, member, and stakeholder perspectives

    • Generating research questions and frameworks

  • Legislative Support

    • Hearing prep drafting

    • One-pagers and section-by-section drafting

    • Reformatting memos

  • Communications Support:

    • Drafting initial versions of newsletters

    • Creating social media content ideas

    • Developing talking points frameworks

    • Editing and improving writing clarity

    • Conceptual translation for different target audiences

  • Administrative Efficiency:

    • Meeting agenda creation

    • Email template development

    • Event planning assistance

    • Training material development

    • Language translation

  • Constituent Services:

    • Developing FAQ responses (using public information only)

    • Creating resource guides for common issues

    • Drafting template responses (without PII)

  • Workflow Automation:

    • Batch file processing

    • Data analytics

    • Local microsite deployment

    • Scheduling daily routines

This guide was developed by POPVOX Foundation to support Congressional staff in using AI tools responsibly and effectively.

About POPVOX Foundation

POPVOX Foundation is a 501(c)3 nonpartisan nonprofit organization with a mission to help democratic institutions keep pace with a rapidly changing world. Through publications, events, prototypes, and technical assistance, the organization helps public servants and elected officials better serve their constituents and make better policy.

Questions or feedback? Contact us at info@popvox.org

Aubrey Wilson
Next

Open, Closed, and Customizable: A Primer on Model Selection