Comparing AI Adoption in the US Congress to Legislatures Worldwide
BY MARCI HARRIS
Despite initial momentum, the US Congress risks lagging international counterparts in the adoption of automated technologies. While the House and Senate have issued guidance allowing some experimentation with AI, the current approach is uncoordinated and piecemeal in comparison to the integrated strategies employed by legislatures such as Norway's Storting, Estonia's Riigikogu, Chile’s House of Deputies and the European Parliament, which are leveraging AI for real-time transcription, legislative process improvements, and data-driven workflows. To improve its own operations and keep pace with international counterparts, Congress should work collaboratively to develop a cohesive AI strategy, provide dedicated resources for staff and Member training, remove barriers to the use of commercial and open source tools, and engage more actively in global knowledge-sharing.
Recent reports from the Inter-Parliamentary Union (IPU) and the Law Library of Congress chronicle how parliaments worldwide are leveraging AI in their operations. However, these analyses notably exclude an assessment of the US Congress, leaving a critical gap in understanding how one of the world's most influential legislative bodies is navigating this paradigm shift.
Innovative technology in legislatures in selected countries
Report from the Global Legal Research Center of the Law Library of Congress.
The GLRC is a division of the Law Library of Congress staffed with foreign-trained lawyers and research analysts from around the world. Reports by the Global Legal Research Center are available on Law.gov.
The IPU’s World e-Parliament Report 2024
Published by the Inter-Parliamentary Union (IPU).
The report examines how parliaments worldwide use technology to improve transparency, efficiency, and public engagement. It highlights trends like AI adoption, open data, and digital transformation, providing a global benchmark for legislative innovation.
Our analysis aims to fill the gap by providing an overview of the current state of AI adoption in the US Congress, highlighting specific initiatives within key legislative support agencies, and comparing these efforts with those of international counterparts. By examining how other legislatures are integrating AI into their legislative processes, we identify opportunities and strategies for Congress to responsibly accelerate its own AI strategy.
Current State of AI Adoption in the US Congress
-
In many ways, the US Congress was an early mover on algorithmic modeling and adoption of automated technologies. This includes predictive modeling by the Congressional Budget Office and Joint Tax Committee since the 1980s, early experiments with machine learning and natural language processing in the Government Accountability Office (GAO)’s Innovation Lab, and the House’s 2022 launch of the Comparative Print Suite, which uses natural language processing to compare bill versions and present visualizations of potential amendments. Following the launch of ChatGPT and the subsequent rise in awareness of generative AI (GenAI), in 2023 the Committee on House Administration (CHA) engaged a GAO detailee to begin working on an AI policy for the House. Around the same time, the House Digital Service distributed 40 ChatGPT Plus licenses to a bipartisan staff working group to collect real-world use cases and feedback about how AI could enhance Congressional operations.
In September 2023, the Committee on House Administration (CHA)’s Modernization Subcommittee released its first “Flash Report,” publicly documenting the progress of AI initiatives throughout the Legislative branch, and the Chief Administrative Officer (CAO) issued preliminary guidance for AI use. This early guidance authorized ChatGPT Plus for research and evaluation use with non-sensitive data if privacy settings were enabled. CHA also released “AI Guardrails” emphasizing the need for human oversight, clear policies, robust testing, transparency, and staff education for further AI adoption in the House.
Despite the initially proactive approach to allowing experimentation with emerging AI tools in the House of Representatives, the House Information Technology Policy 08.0 (HITPOL 08.0) released in September 2024, establishes a new restrictive framework with CHA/CAO as gatekeeper through which Members and staff must proactively seek pre-approval for new uses or tools and over a year since the initial guidance was released, ChatGPT Plus is still the only commercial AI tool approved for use in the House.
In our view, the new policy — if followed by lawmakers and policy staff — would significantly curtail their ability to fully understand the tools and technologies that they are endeavoring to regulate in their policy roles. An important feature of the HITPOL 08.0 policy is that it is slated to be reviewed and revised annually, providing an opportunity to correct the overly restrictive language of the initial policy.
-
In December 2023, the Senate Sergeant at Arms issued guidance that embraces a more permissive strategy than the US House, with a focus on experimentation and learning. Rather than restricting access to a single tool, the Senate authorized multiple AI platforms available at the time, including OpenAI's ChatGPT, Microsoft's Bing Chat AI, and Google's Bard. The Senate's framework prioritizes building staff familiarity with AI technologies while managing risks through oversight rather than restriction. However, the Senate’s guidance has not been updated since the initial release and many provisions (including significant changes to the approved platforms — Microsoft’s AI tool is now “Copilot” and Google’s tool is now “Gemini”) mean that a revision is overdue.
AI Initiatives Across Institutional Offices and Legislative Support Agencies in the US Congress
The summaries below are based on the "Flash Reports" published by the Committee on House Administration's Subcommittee on Modernization in 2023 and 2024, public reporting from the agencies, and interviews with staff.
-
Governance: Established an AI Advisory Committee and AI Power User Group for oversight. Implemented a new AI solution assessment process.
Use Cases: Considering Microsoft Copilot and Azure’s OpenAI platform for pilot programs. Conducting a quality management pilot with AI-powered robotic process automation (RPA).
Staff Upskilling: Provided on-demand AI training and prompt engineering training.
-
Governance: Released the HITPOL 08.0 AI Policy for the House, establishing governance and responsible use principles. Established a workflow for new AI use case requests.
Use Cases: Testing Microsoft Copilot and AWS Bedrock for potential use cases.
Staff Upskilling: Launched internal resources for House staff, including FAQs and a list of vetted AI use cases.
-
Governance: Developed an internal AI policy and AI advisory committee.
Use Cases: Indexed over 16,000 files for an AI chatbot. Developed an advanced AI search tool. Working on an AI procurement plan.
Staff Upskilling: Provided advanced AI training to IT developers. Setting up an AI laboratory.
-
Use Cases: Assessing AI for suggesting bill language, spotting inconsistencies, and flagging legal conflicts.
-
Governance: Established an AI Working Group to document AI use, explore tools, and connect AI innovation with Library strategy. Created an AI Community of Practice.
Use Cases: Conducting experiments with AI for tasks like bill summary automation at the Congressional Research Service.
Staff Upskilling: Offering staff training on AI tools, practices, and ethics.
-
Use Cases: Developed the Comparative Print Suite in collaboration with the HOLC to improve legislative functions.
Focus: Looking into specific AI applications that enhance legislative processes.
-
Use Cases: Experimenting with AI to automate legal analysis, review statutes, and identify inconsistencies.
Partnerships: Working with external vendors to pilot new approaches.
-
Governance: Developing an AI Charter and formal policy directive to govern AI use.
Use Cases: Integrated AI into security operations, focusing on threat analysis and protective operations.
Collaboration: Seeking advice and guidance from sources like the National Policing Institute.
-
In January 2024, the GAO reported on its active use of AI to improve efficiency and oversight, including deploying a customized large language model for synthesizing reports and scanning congressional mandates. Among eight use cases under development are automated text organization, legislative mandate summarization, copyediting, and real-time congressional information monitoring. These efforts align with GAO's broader AI Accountability Framework, ensuring responsible use while enhancing internal processes and providing Congress with technical assistance and oversight capabilities.
-
Based on interviews with staff from the Congressional Budget Office (CBO), we know that the agency has has explored AI tools for its operations with a focus on coding, legislative analysis, and process improvement. CBO has experimented with Microsoft Copilot to translate legacy Fortran code into R, expanded AI licenses for broader staff adoption, and initiated a project to parse legislative text for expiring appropriations tracking. An internal AI Working Group of 20 analysts is exploring the potential use of AI for data standardization and tracking expiring program authorizations.
Global Legislative AI Adoption: A Comparative Perspective
As countries adopt AI to improve efficiency, transparency, and responsiveness, each presents unique approaches to overcoming challenges. This section examines key AI adoption initiatives in several legislatures — Norway, Estonia, Chile, India, and the European Union — and draws comparisons to the US Congress, highlighting lessons that can be applied to strengthen US legislative AI adoption.
Norway: The Digital Value Chain Initiative
Norway's Storting (parliament) is implementing one of the world’s most ambitious AI-driven initiatives, the "Digital Value Chain" project, slated to launch in 2025. The “Storsak” initiative aims to automate the legislative process, streamlining drafting, reviewing, and debating legislation by moving from document-based to data-first workflows. The system uses machine learning (ML) to manage real-time data exchanges between the government and parliament, tagging documents with relevant metadata to make searching and retrieval seamless. The project takes a holistic approach to AI integration in the legislative process, with a goal of transitioning “from documents to data.”
Lesson for US Congress: Norway’s comprehensive use of AI to digitize legislative processes underscores the potential for a strategic, coordinated approach. Congress could explore similar data-first workflows, which would move beyond pilot programs and toward a more integrated and automated system for legislative data management. Such a shift could reduce reliance on manual document handling and provide real-time access to legislative data, improving decision-making speed and accuracy.
Related: On June 21, 2024, the offices of the Clerk and the Legislative Counsel issued a Request for Information (RFI) seeking input on technology, architecture, and process solutions related to collaborative legislative drafting. This initiative, known as the "Collaborative Legislative Drafting Study," marks a pivotal moment in the ongoing efforts to enhance the efficiency and effectiveness of lawmaking in the digital age.
Estonia: AI-Powered Speech Recognition and Legislative Processing
Estonia's legislature (the Riigikogu) uses AI for real-time transcription of parliamentary sessions and intelligent search within legal documents. The AI-based system, HANS, transcribes debates with a 5% error rate, reducing documentation time from hours to just 20 minutes and has led to reassignment for the team of stenographers that used to produce transcripts manually.
The Riigikogu is exploring AI-generated committee meeting summaries based on HANS transcripts, while emphasizing the need for human oversight to ensure accuracy and neutrality. Additionally, the parliament's legal department employs an AI-powered search tool developed by Nortal to quickly search through legislative documents, including EU directives, significantly reducing manual research time.
Lesson for US Congress: Estonia's use of AI for transcription and intelligent search shows the potential for automating similar processes in Congress. AI-driven transcription could improve access to legislative proceedings, while intelligent search tools could enhance legal research efficiency. The Riigikogu's experience highlights the importance of balancing innovation with human oversight and ongoing discussions about AI requirements and limitations.
Chile: The CAMINAR Project
Through its “CAMINAR” project, Chile has taken a strategic and phased approach to integrating AI into legislative and administrative functions within its Chamber of Deputies. As Luís Kimaid of Bussola Tech has reported: “The legislative modules of CAMINAR integrate LLMs from OpenAI ChatGPT, and Claude with RAG (Retrieval-Augmented Generation).” CAMINAR-L (Legislation) includes several components:
CAMINAR-L2: Semantic Search Assistant
CAMINAR-L4: Legislative Tracking Assistant
CAMINAR-L5: Regulatory Impact Assessment Assistant
CAMINAR-L6: Constitutional Support Assistant
CAMINAR-L8: Legal Doctrine Assistant
CAMINAR-L9: Management of Budget Amendments
CAMINAR-A (Administration) modernizes administrative tasks, including expense management and financial oversight, by leveraging AI to automate data extraction and compliance checks. By maintaining a balance between automation and human validation, CAMINAR ensures accuracy while fostering trust in AI-driven processes.
Lesson for US Congress: CAMINAR demonstrates that a phased approach, starting with targeted pilot projects and gradually expanding based on proven outcomes, can mitigate risks while building institutional support. By aligning technology deployment with clear goals, such as improving legislative drafting and administrative efficiency, Congress can modernize operations without overwhelming existing structures. Emphasizing human oversight and stakeholder engagement, as CAMINAR does, would ensure that AI integration complements rather than disrupts core functions.
India: Real-Time Translation and Multilingual Support
In India, AI has become essential in overcoming language barriers in the Indian parliament, where Members of Parliament (MPs) speak multiple languages. AI-driven real-time translation tools enable MPs to debate in their native languages, with instantaneous translations available to other MPs. This system has increased participation and accessibility, ensuring that language is no longer a barrier to engagement in legislative sessions.
The AI-powered translation system is not only used during live sessions but also for public documents, ensuring that legislative materials are accessible to constituents in their preferred languages.
Lesson for US Congress: While the US Congress does not face the same linguistic diversity challenges as India, AI-driven translation tools could improve accessibility for non-English-speaking constituencies. AI-powered translation services could also help US lawmakers engage more effectively in multilingual communities and provide legislative materials in multiple languages, increasing transparency and civic participation.
European Union: Comprehensive AI Integration for Multilingual and Legislative Efficiency
The European Parliament has taken a comprehensive approach to AI integration, deploying across several functions, from multilingual support to enhancing archive accessibility. Notably, AI is used to manage multilingual communications in a legislative body where members speak over 20 languages. AI-driven speech recognition tools are used for transcription, and AI supports the translation of parliamentary sessions and documents into multiple languages, ensuring that debates and legislation are accessible across the EU's diverse linguistic landscape.
Recently, the European Parliament transformed the accessibility of its archives by leveraging Anthropic's Claude AI, creating “Ask the EP Archives” or “Archibot.” This AI assistant makes over 2.1 million official documents from 1952 onwards readily available to researchers, policymakers, educators, and the public. Archibot offers users worldwide easy access to EU Parliament documents in multiple languages, including resolutions, positions, policies, and inter-institutional negotiations. It also helps researchers and staff analyze data and create comprehensive reports.
Key features of Archibot include:
Advanced search and summarization, allowing users to navigate and synthesize vast amounts of information quickly
Report building, enabling users to extrapolate information and build reports
Multilingual capabilities, expanding language support beyond French to serve all EU member states
Global accessibility through the European Parliament's website
Constitutional AI, ensuring trustworthy and controlled use of AI technology
The project has already demonstrated significant benefits, reducing document search and analysis time by 80%, providing policymakers and researchers with quick access to historical context and precedents, and being used by hundreds of educators and students for an all-access window into European parliamentary history.
Lesson for US Congress: Lesson for US Congress: The European Parliament's "Ask the EP Archives" project, powered by Anthropic's Claude AI, showcases the potential for transforming legislative archive accessibility. By leveraging Claude's advanced natural language processing, the EU has made over 2.1 million official documents easily searchable and accessible, reducing search time by 80% and providing multilingual access to historical information.
In contrast, the US Congress currently prohibits Members and staff from using platforms like Claude, hindering the adoption of AI for enhancing archive accessibility. Congress could learn from the EU's example and explore AI-powered tools to make its vast legislative archives more user-friendly, efficient, and transparent.
To achieve this, Congress would need to address current restrictions on AI platform usage and collaborate with providers to develop a secure, tailored solution. By doing so, Congress could modernize its archive management and make its rich history more accessible to the American people, following in the footsteps of the European Parliament's successful implementation of Archibot.
What the US Congress Can Learn from AI Adoption in other Legislatures
Comprehensive AI Strategy and Leadership
Norway's coordinated "Digital Value Chain" project demonstrates the value of a unified and comprehensive approach to AI adoption. By contrast, the US Congress has adopted a more piecemeal approach, with each chamber and legislative support agency experimenting with AI independently. Establishing a more cohesive AI strategy for Congress — potentially led by a dedicated office or task force — could accelerate adoption and ensure consistency across different parts of the legislative process.
Automated Transcription and Documentation
Both Estonia and Chile have successfully implemented AI-powered transcription services, reducing the time and labor involved in producing parliamentary records. The US Congress, which still relies heavily on manual transcription, could implement similar AI tools to create real-time records of House and Senate proceedings, improving both efficiency and transparency.
Intelligent Search and Legal Document Management
Estonia's use of AI for intelligent search within legal documents showcases the potential for AI to assist legislative counsel and researchers. The US Congress, particularly the House and Senate Legislative Counsel offices, could benefit from AI-driven search tools that reduce the time spent manually reviewing legal texts and drafting legislative language.
Data-First Legislative Processes
Norway’s shift toward a data-first, metadata-driven legislative system offers a model for Congress to streamline workflows and reduce manual effort. Congress could adopt similar methods, moving away from document-based systems and toward integrated, automated processes that ensure data consistency and transparency across all legislative activities.
Translation and Multilingual Accessibility
While the US does not face the same linguistic diversity challenges as India, AI-driven translation tools could help Congress better serve multilingual constituencies and enhance international collaboration by providing real-time translations of legislative materials and public-facing documents.
The US Congress is making progress in its AI adoption efforts, but global counterparts offer valuable lessons on how to accelerate and expand these initiatives. Norway’s data-first approach, Estonia’s real-time transcription and intelligent search tools, Chile’s automated legislative tracking, India’s real-time translation system, and the European Union’s multilingual AI integration all provide models that Congress can learn from.
By studying these international examples, Congress has the opportunity to adopt AI more comprehensively to enhance legislative processes, improve transparency, and increase accessibility. Embracing a broader AI strategy that encompasses legislative drafting, research, and constituent engagement will allow Congress to maintain its global leadership in governance and ensure it stays competitive in the rapidly evolving digital landscape.
