2025 Maryland AI Enablement Strategy & AI Study Roadmap

​​​​​​​​​​​​​​MEMORANDUM FOR: Governor Wes Moore and the Maryland General Assembly
FROM: The Governor’s AI Subcabinet
SUBJECT: 2025 Maryland AI Enablement Strategy & AI Study Roadmap​

Executive Summary​

• The State of Maryland is on a journey to build its capacity to responsibly, ethically, and productively leverage AI to increase government effectiveness and improve constituent outcomes.
• This memo outlines the State’s high-level 2025 AI Enablement strategy and, as required by the AI Governance Act of 2024 (SB818), includes plans for studying opportunities, risks, and next steps associated with the use of AI in State services in a variety of key domains.
  

AI Enablement Strategy

​​​​AI Enablement refers to the State’s efforts to ensure its components are set up for success in leveraging AI and Machine Learning (ML) technologies to improve constituent outcomes and managing/guiding these technologies’ myriad potential halo effects, with the capabilities to do so responsibly, ethically, and productively. AI is not a panacea or short term fix for systemic issues, but it is a rapidly evolving new platform technology with real potential as a critical tool in agency quivers. To that end, this work is a collaboration between many bodies, including the AI Subcabinet, the Governor’s office, the Department of Information Technology (DoIT), and working teams across many agencies.

2024: Foundation-building​

​​​​The Governor’s AI Executive Order was signed in January 2024, and the AI Governance Act (SB818) was passed and made effective as of July 2024. These pieces of policy kicked off th​e State’s work in earnest. Since then, the AI Subcabinet, Department of Information Technology (DoIT), and stakeholders across the State began cohering foundational building blocks to increase the State’s capabilities.

​​​​We can split these foundational activities into the “responsible” and “productive” use of AI. Under “responsible use,” this included disseminating interim GenAI guidelines; collecting the first state-wide AI inventory; making available free AI training​ for the state workforce; and crafting a “v1” state-wide AI intake process. Under “productive use”, this included running initial proofs-of-concept (PoCs) to start “learning by doing” (examples include PoCs at the Department of Environment, MDThink, and MD’s Open Data Portal); kicking off a state-wide AI Community of Practice; building partnerships to provide the state with expertise and advisory 3 support; and hiring for a new AI Enablement team at DoIT to get in place relevant expertise and “incubate” initiatives.

​​​​A number of agencies also set up their own working groups and innovation teams, in order to think through domain-specific policy, adoption, and use cases, and experimented with commercial-off-the-shelf GenAI tools like ChatGPT, Claude, and Gemini. Throughout, we engaged with industry, civil society, academia, and federal/county/municipal governments across Maryland, the US, and abroad to ensure our approaches are rooted in evolving best practices, without re-inventing the wheel.

​​​​2025: Experimentation and momentum-building

​​​​In 2025, we will build momentum on these efforts, clarify operating models, and in particular, seek to increase the pace of experimentation, iteration, and adoption. Our high-level strategy can be encapsulated as follows.

1. Mature the State’s AI governance capabilities

• ​Current use of AI in the State is relatively minimal, and existing risk management processes have been sufficient as the de facto approach. However, as the volume of AI requests and solutions in use and production increase in 2025 onwards, the unique risk management needs for AI systems during intake, procurement, deployment, and monitoring will increase in tandem.
• Ensure adherence to the principles outlined in the Governor’s AI EO, as well as existing legal and policy requirements, by operationalizing a more formal AI governance framework and associated artifacts via the State’s software intake and procurement processes. We will streamline and integrate that framework into existing risk management processes and iterate from a minimum viable “v1” version to greater maturity over the year.
• Agencies will build more sector-specific guidelines on top of, or integrated within, this process as needed.

2. Strengthen the State’s data foundations

• ​A strong data foundation is the core requirement for effective AI experimentation, adoption, and governance, with trusted and reliable outputs.
• Formalize the State’s data governance processes and ensure agencies have reliable data for testing and scaling AI solutions. This includes ensuring agencies have the tools and resources necessary to test risks associated with using datasets for AI, providing alternative datasets (i.e., synthetic data) to protect the security and privacy of state data, and cataloging data that is allowed/restricted for use with AI solutions.
• Establish an Authoritative Data Sources program to prioritize critical state datasets, outline the standards for “AI-readiness”, and ensure identified datasets meet those standards, in order to power experimentation and adoption.

3. Build momentum around experimentation & adoption

• ​Move from the more “opportunistic” experimentation approach of 2024 to a more structured approach, leveraging clearer governance processes, evaluation frameworks, a newly formed state AI Enablement Team, clearer agency goals around leveraging AI, and fit-for-purpose procurement pathways.
• Establish and execute a variety of experimentation channels across the State to expedite learning and time-to-impact. These may range from pilots of commercial-off-the-shelf (COTS) AI tools to determine return-on-investment (ROI), common use cases, and areas for future investment; to focused hackathons; to a portfolio of more structured Proof-of-Concept (PoC) procurements focused on specific, prioritized, high-impact problem statements; to workshops helping agencies understand what is possible with AI and design prototypes; to a statewide sandbox environment to encourage more decentralized experimentation.
• Build clear playbooks to help agencies move from successful experiments to scaled solutions in production, including greater clarity on build vs. buy decisions and avoiding lock-in.
• In parallel, evaluate options around technical infrastructure, architecture, and platform strategy to underpin future growth.

4. Increase the state’s “AI IQ”

• ​Establish and deepen programs and partnerships that increase the State’s AI literacy, talent, and available expertise.
• Increase the percentage of state workers taking AI courses the state makes available, and expand the types of courses available based on the broader workforce strategy noted in the critical domain deep dives.
• Launch a Maryland Data Academy to create opportunities for state employees to better understand how data influences and drives AI solutions. The Academy will provide opportunities for employees to engage with use cases, participate in a variety of training courses, and have access to a set of resources to support scaling the use of data responsibly.
• Build low-friction mechanisms for students and researchers at Maryland academic institutions to serve as interns, fellows, or short-term experts on agency AI projects, in order to increase opportunities for Maryland students while benefiting state services.
• Strengthen and expand the State’s AI Community of Practice as a collaboration and upskilling mechanism across state, county, and local government.

5. Study and cohere state approaches to AI in critical domains

• ​​Develop an understanding of risks, opportunities, and approaches in key topic areas including those put forth by SB818.
• In aggregate, these studies and their outputs should provide a solid foundation to ensure the State is approaching AI-related impacts with a strong grounding in Maryland-specific trends and constituent perspectives.

​​​​Study Topics & 2025 Roadmap

​​​​As mandated and outlined by SB818, and to execute the “critical domains” portion of the State’s 2025 strategy, state agencies, together with relevant internal and external stakeholders, will study the topics noted below through 2025, with deliverables to be shared by December 2025 with the Governor and General Assembly. Outputs of these studies will depend upon findings, topic context, and resourcing but may include reports, prioritized recommendations, pilot initiatives, agency workstreams, or changes in policy.​

​​​​Workforce

​​​​Relevant SB818 requirements:
“A plan to study the use of AI by the State workforce, including opportunities to upskill the workforce.”

​​​​“A plan to study the use of AI in workforce training.”

​​​​“A plan to study the hiring of talent with expertise in artificial intelligence, employment practices, and workforce development implications.”

​​​​The state government workforce

​​​​The State’s goal is to integrate AI and train its workforce in ways that enhance, rather than replace, human workers - while increasing the efficiency and efficacy of government service delivery. As outlined in the above 2025 AI Enablement strategy, the State will continue experimenting with GenAI tools in state employee workflows, and invest in tools where there is significant ROI and opportunities to improve constituent outcomes and reduce workplace drudgery. We will also build on existing efforts, such as the State’s partnership with InnovateUS, which provides free asynchronous AI trainings and workshops available to all MD state employees.

• Plan:
• ​Collect and analyze industry analysis data to determine what public sector jobs will require which types of additional AI skills.
• Meet with labor leaders and organizations to incorporate worker perspectives.
• Clarify areas where AI should not be used or deployed in a more limited way.
• Collect information about existing and emerging AI skills training programs to determine which meet the needs of the State.
• Continue identifying and experimenting with use cases that can improve workflow and efficiencies and find the best routes to adoption.
• Engage with federal bodies that have been successful in hiring AI talent, such as the DHS AI Corps, to adopt best practices for prioritized roles.
• Leverage existing research, engage with experts, and outline where AI can be responsibly, ethically, and productively leveraged in State hiring, recruitment, and onboarding processes.
• Stakeholders
• Leads:
• Dept. of Labor (MDL)
• Dept. of Budget & Management (DBM)
• Involve:
• ​Dept. of Information Technology (DoIT)
• MD Higher Education Commission (MHEC)
• Labor unions
• Relevant civil society organizations and academic researchers focused on AI workforce
• Relevant technology companies and hyperscalers with insight into workforce implications
• Civil Rights Commission
• Projected Timeline:
• By June 2025:  Conduct industry and skills analysis
• By Sept. 2025:
• Gather information from state agencies on occupations identified in industry and skills analysis and also work with public sector labor representatives to determine the best way to collect information and identify potential impacted occupations.
• Conduct partner outreach and discussions.
• By Sept. 2025:  Analyze results and share recommendations.

​​​​The broader state workforce

​​​​In the coming decade, AI is likely to impact Maryland’s workforce in predictable and unpredictable ways. These impacts will manifest differently by industry, job type, demographic, and location. Our work seeks to develop the baselines from which to encourage promising 7 trends, mitigate downside risks, and ensure the availability of programs whereupon the workforce has tools to upskill and thrive.

• Plan:
• Conduct an analysis to identify industries, skills, and occupations that are prevalent in Maryland that are at greatest risk of being impacted and those that are most likely to require workers to develop new or enhanced AI-related skills over the next 3-5 years.
• Meet with labor leaders and experts to incorporate worker voices into AI policy analysis and determine needs around AI training programs.
• Meet with groups representing underserved communities to incorporate their voices into AI policy analysis and determine needs around AI training programs.
• Stakeholders
• Owners:
• MDL
• MHEC
• ​Involve:
• Dept. of Commerce
• Labor Unions
• Community Colleges
• Universities and academic researchers
• Projected Timeline:
• ​By June 2025:  Conduct industry and skills analysis.
• By Sept. 2025:  Conduct partner outreach (use labor market analysis to inform discussions).
• By Dec. 2025:  Analyze survey results and partner input and share recommendations.

​​​​Critical Infrastructure

​​​​Relevant SB818 requirement:

​​​​“A plan to study the use of AI in critical infrastructure, including guidelines for owners and operators to incorporate risk management into critical infrastructure.”

​​​​The opportunities stemming from leveraging AI in critical State infrastructure, and the new attack surfaces AI engenders, requires a highly Maryland-specific plan that is based on the State’s unique context, institutions, and systems - ranging in diversity from our energy systems, to our waterways, to our food and agriculture sector. As the foundation for this work, we will leverage the “Roles and Responsibilities Framework for AI in Critical Infrastructure” document, developed by the Department of Homeland Security in consultation with its AI Safety and Security Board (AISSB), a body for which Governor Moore served as a Board member throughout 2024.​

• Plan:
• Review the AISSB framework, conduct a broader literature review, and interview internal and external experts.
• Explore creating a state-level AI Critical Infrastructure Safety and Security Advisory Council to assist with the work, provide direction, and review recommendations and outputs.
• Determine if state-level guidance is necessary for various critical infrastructure sectors.
• Identify potential pilots of tools, programs, or initiatives.
• Stakeholders
• Owners:​ ​Governor’s Office of Homeland Security (GOHS)
  
• Involve​:​
• ​MD Dept. of Emergency Management (MDEM)
• MCAC
• MD Dept. of Transportation (MDOT)
• MD Energy Administration (MEA)
• DoIT
• Owners and operators of key identified infrastructure
• Projected Timeline:
• ​By March 2025:
• ​Review the AISSB framework and other existing guidance and identify key actions for the State to take
• Explore the possibility of convening a CIKR AI Safety and Security Advisory Council within the state
• ​By March 2025: Develop a resource library available for CIKR
• ​By March 2025: Develop guidelines​

​​​​Local School Systems​​

​​​​Relevant SB818 requirement:

​​​​“A plan to study the use of AI in local school systems, including recommendations to the State on the responsible and productive use of artificial intelligence.”

​​​​Better understanding and guiding the use of AI in local school systems is one of the state’s most pressing domains, given that the use of AI tools by students, teachers, and administrators is already widespread and rapidly evolving (technologically and normatively). Further, the “AI in workforce upskilling” question cannot be effectively tackled without understanding trendlines in K-12 education. In 2024, the MD State Dept of Education (MSDE) shared an AI Cybersecurity policy with local school systems; created an AI Education Committee to research and draft AI guidance documents; launched a website to share AI resources and provide information to stakeholders; and joined the TeachAI community to collaborate with other states. 2025 work will build on this, addressing tactical concerns and engaging with more philosophical questions around the role of AI in education and the types of skills that might be needed for the future workforce.

• Plan:
• ​Meet with other states and educator communities of practice working to leverage, govern, and secure AI in the classroom and compile emerging literature, studies, case studies, and best practices.
• Conduct and analyze stakeholder engagement and surveys in order to understand the evolving needs of local school systems and the gaps therein.
• Iterate existing policy and provide new guidance and training programs based on conducted research, particularly focused on the selection and integration of AI products into institutional pedagogy.
• Where possible, leverage and build upon more general risk management guidance, processes, and tools from the AI Subcabinet as they are released over the year.
• Stakeholders
• Owners: MSDE
• Involve:
• TeachAI
• MD Center for Computing Education (MCCE)
• Computer Science Teacher’s Association of MD (CSTA-MD)
• Federal Dept. of Education
• Civil society orgs working on AI in education
• Academic researchers
• ​ Projected Timeline:
• Fall 2025: Share the MSDE AI Instructional guidelines
• By Dec. 2025: Deliver learnings and recommendations.
• Winter 2026: Deliver learnings and recommendations.

​​​​Healthcare and Human Services Delivery

​​​​Relevant SB818 requirement:

​​​​“A plan to study the use of AI in health care delivery and human services.”

​​​​AI has transformative potential in State healthcare and human services provision - from disease monitoring and outbreak prediction, to reducing systems’ administrative costs, to channels for constituents to more quickly and intuitively get access to critical information on benefits and vital documents. In 2024, HHS agencies conducted a number of proofs-of-concept in internal-facing, low-risk domains, and that type of experimentation will continue. However, given the sensitivity of the data leveraged in this space for many use cases, and the impact if that data were to be compromised, the lion’s share of focus in these studies will be on robust AI risk management and governance. For example, ensuring that nuances between the different disciplines in this broad area - public health, behavioral health, Medicaid, human services - are accounted for in AI risk management and procurement processes.

• Plan:
• Establish v1 AI policy and guidelines to encourage AI solutioning for MDH health services delivery challenges.
• Explore HHS-specific data and GenAI risk management and mitigation strategies, building on top of broader policies, processes, and requirements for AI governance established by the AI Subcabinet.
• Explore the use of generative AI for measurement of health services delivery to MD constituents.
• Assess and conduct surveys and interviews on areas where AI may be able to enhance constituent-facing service delivery.
• Assess MDThink GenAI proofs-of-concept and, if relevant, develop adoption roadmap and applicability to other HHS use cases.
• Stakeholders
• Owners:
• MD Dept. of Health (MDH)
• MD Dept. of Human Services (DHS)
• Involve:
• MDThink
• MHA (MD Hospital Association)
• Relevant federal, state and local government Agencies
• Academics focused on this area
• Vendor community
• IT Advisory Organizations
• ​DoIT
• Projected Timeline:
• By March 2025: Publish sector-specific AI guidelines
• By June 2025: Publish sector-specific AI guidelines
• By Sept. 2025: Conduct survey
  
• By Dec. 2025: Compile and publish findings + recommendations​

​​​​Cybersecurity

​​​​Relevant SB818 requirement:

​​​​“A plan to study the use of AI in the discovery and remediation of vulnerabilities in cybersecurity and data management across State and local government, including school systems.”

​​​​We will study two lenses in this domain: “cybersecurity for AI”, and “AI for cybersecurity”. The former refers to risks that AI tools pose, both in the course of use by the State (eg - data loss), and in their use by adversaries to underpin more sophisticated attacks on the State (eg - spearphishing). The latter refers to leveraging AI tools to yield better outcomes in existing cybersecurity workstreams. The results will inform iterations to our existing cybersecurity processes, data management approaches, software intake flow, and potential pilots, and build on the AI and data inventories conducted in 2024.

• Plan:
• Via literature review and discussions with internal and external experts, peers, and partners, establish or adopt a framework that categorizes cybersecurity risk factors of different types of AI solutions, maps them to relevant mitigations, and recommends methods to operationalize them in the State’s AI and data governance processes.
• Integrate these suggestions into MSDE’s existing AI Cybersecurity policy as warranted.
• Meet with cybersecurity leadership across the State, counties, and municipalities to survey their organizations on areas where AI can be a force multiplier and where they might need assistance from the State.
• Partner with internal stakeholders and external experts in industry, government, and academia to create a prioritized list of promising use cases, mapped to existing friction points, of leveraging AI for cybersecurity and data management (for example, automating the detection of vulnerabilities and anomalies in how data is discovered, accessed, and used, or automating data integrity and quality evaluation).
• Identify and execute proofs-of-concept to test utility for prioritized use cases above.
• Stakeholders
• Owners:
• DoIT OSM
• Involve:
• ​MSDE
• DoIT Office of Enterprise Data
• MHEC
• MDK
• 12 committee
• Academic researchers
• Industry SMEs
• ​ Projected Timeline:
• By Dec. 2025: Completed cybersecurity risks framework, prioritized use cases, and relevant pilots.​

​​​​Economic Development

​​​​Relevant SB818 requirement:

​​​​“A plan to study the use of AI to support job and business creation and growth in the State.”

​​​​The AI revolution has potential as a force multiplier for the State’s economic development goals. Commerce already engages around AI with commercial stakeholders in the business community from various sectors and has identified a pressing need for baselining the work in MD-specific trends, perspectives, and potential scenarios. This will help the State understand current trendlines, metrics that could measure and signify success, goals to aim for, and better formulation of policies and initiatives that would help meet those goals.

• Plan:
• Conduct an economic impact analysis focused on job creation.
• Run industry surveys, interviews, and workshops with business leaders and entrepreneurs to better understand needs and ways to support job and business creation and growth.
• Conduct and review case studies in other geographies.
• Workshops with business leaders and entrepreneurs.
• Stakeholders
• Owners:
• Dept. of Commerce
• Involve:
• Chamber of Commerce
• Industry Groups
• University researchers
• Department of Labor
• Workforce development commissions & agencies
• Projected Timeline:
• ​By April 2025: Conduct discussion sessions
• By Dec. 2025: Deliver outputs (economic impact report, refined existing plan, prioritized recommendations)​

​​​​Data Privacy

​​​​Relevant SB818 requirement:

​​​​“A plan to study the use of AI in data privacy, specifically regarding the ability to train systems that employ artificial intelligence.”

​​​​The 2024 AI Executive order establishes Privacy as one of the State’s core principles governing AI adoption and is a key component of our existing evaluation process. Along with strong data governance and management practices, robust privacy tools, processes, and policies are a necessary condition for the broader adoption of AI solutions. We will leverage a deeper understanding of emerging best practices in privacy & AI to mature and evolve our AI governance practices.​

• Plan:​
• Surface and prioritize state actions, policies, practices, and mechanisms to strengthen privacy protections for state and constituent data used in AI-enabled systems to align with the Fair Information Practice Principles.
• Leverage state data classification policy to draft and operationalize privacy policy outlining the acceptable and unacceptable use of confidential State data to train AI systems.
• Create a risk register comprehensively outlining types of privacy risks found at different points in current or anticipated State AI deployment processes. Determine mitigations which include specific timelines and tasks to address said risks and approaches to implementing prioritized mitigations in state procurement and operations.
• Determine the role of privacy-enhancing technologies (PETS) (e.g., encryption, anonymization, differential privacy, homomorphic encryption) in protecting personally identifiable and confidential information in the context of the state’s use of AI, by conducting comprehensive literature reviews, getting input from internal and external SMEs, and inviting PET vendors in to demonstrate their tools.
• Redesign, develop, and deploy privacy training that incorporates AI best practices for authorized users of PI.
• Stakeholders
• Owners:
• DoIT (Privacy, OSM, Office of Enterprise Data)
• Involve:
• DGS
• DoIT AI Enablement
• OAG
• External SMEs (relevant think tanks, civil society orgs, industry)
• ​ Projected Timeline:
  
• By June 2025:
• Prioritization
• Acceptable Use
• By Oct. 2025:
• Risk register
• By Dec. 2025:
• Recommendations
• Training​

​​​​Criminal Justice

​​​​Relevant SB818 requirement:

​​​​“A plan to study the use of AI in the criminal justice system, including whether and how such technology should be used, in what contexts, and with what safeguards.”

• Plan:
• Create a subcommittee of the Maryland Judicial Council tasked with developing internal and external guidance for the responsible use of AI.
• Assess and conduct surveys and interviews with relevant internal and external stakeholders on areas where AI may be able to enhance public-facing judicial services.
• Study and prototype AI process automation to improve clerk functions, and the deployment of chatbots that facilitate seamless retrieval of relevant information for individuals seeking insights into MD courts.
• Stakeholders:
• Owners:
• Judicial Information systems
• Involve:
• MD Administrative Office of the Courts
• Circuit Courts
• District Courts
• MD Office of the Public Defender (OPD)
• MD Dept. of Public Safety & Correctional Services (DPSCS)
• Office of the States’ Attorney
• Projected Timeline::
• ​Throughout 2025.​ ​

​​​​Public Safety

​​​​Relevant SB818 requirement:

​​​​“A plan to study the use of AI for public safety purposes, including whether and how such technology should be used, in what contexts, and with what safeguards.”

​​​​AI technologies present both significant opportunities and serious risks in public safety and law enforcement contexts. While AI could enhance capabilities in areas like emergency response optimization, crime analytics, and administrative efficiency, its use raises critical concerns around civil liberties, algorithmic bias, transparency, and community trust. Several high-profile cases nationally have highlighted how AI in law enforcement, if not properly governed, can perpetuate historical biases or infringe on individual rights. Maryland's approach will therefore be deliberate, measured, and transparent, in order to develop a framework that ensures that AI deployment in law enforcement upholds both public safety and civil rights. This includes noting AI systems in use in an AI Inventory, for example, and characterizing their risk level.​

• Plan:
• Meet with state police departments to explore the current and potential use of AI to: a) Collect approaches to and perspectives on developing a framework for leveraging AI in law enforcement, and b) Collect low-risk use cases that can improve public safety outcomes + evaluate potential proofs-of-concept
• Leverage inputs from the above meetings to add law enforcement-specific risk evaluation to the State’s overall AI intake process, specifically for higher-risk tools in public safety that require impact assessments
• Implement an AI Task Force with MCAC to assist law enforcement with the analysis of criminal investigations that involve the use of AI in criminal activity
• Stakeholders:
• Owners:
• MD State Police (MDSP) / MD Coordination & Analysis Center (MCAC)
• Governor’s Office of Homeland Security (GOHS)
• Involve:
• MD Transit Administration (MTA)
• MD Transportation Authority (MDTA)
• MD Natural Resources Police (NRP)
• DoIT/CPO​
• Projected Timeline:
• Throughout 2025.​​

​​State Licensed and Certified Occupations​

​​​​Relevant SB818 requirement:

​​​​“A plan to study the use of AI by occupations licensed and certified by the State, including identifying ways for State regulatory boards to identify and manage the risks and opportunities of AI and determine appropriate permitted use and supervision of licensees.”

​​​​Related to efforts around AI & Workforce, occupations licensed and certified by the State offers an opportunity to work closely with workers in a variety of occupations to understand trends, needs, and approaches to catalyzing responsible and productive AI enablement in their work, as well as exploring the use of AI in the licensing and certification process itself.

• ​ Plan:
• Survey occupational licensing boards to determine if AI is becoming a critical skill within the occupation that should be considered in the licensure of specific occupations.
• In unionized industries, survey labor unions to determine if AI is relevant to licensure in specific occupations.
• In industries with national professional associations, national board councils, and/or national exam proctors, survey these stakeholders to determine if AI is relevant to exam materials, licensure, or industry practices.
• For licenses that require completing continuing education for license renewal, survey continuing education providers to determine how training on the use of AI tools has been or may be used in continuing education courses.
• Survey licensing boards about operations and licensee issues to identify areas of necessary improvement.
• Investigate methods to better monitor licensing process duration and make improvements, as well as provide more effective reporting and automation.
• Stakeholders:
• Owners:
• MDL
• MDH
• Involve:
• Labor Unions
• National Professional Associations
• National Board Councils/Exam Proctors
• Continuing Education Providers
• Occupational Licensing Boards
• IT Advisory Organizations
• Community Colleges
• Universities
• Federal, State and Local Agencies
• Vendor Community
• Projected Timeline:
• By June 2025: Develop a survey and send the survey to stakeholders.
• By Sept. 2025:
  • Analyze survey results.
  • Conduct partner outreach and discussions (use survey results to inform discussions).
• By Dec. 2025: Analyze results and share recommendations.
  

​Conduct of Elections

​​​​Relevant SB818 requirement:

​​​​“A plan to study the use of AI in the conduct of elections, including reducing or eliminating the spread of misinformation.”

​​​​The integrity of elections is fundamental to democratic governance, and AI's role in this space presents both opportunities to strengthen electoral systems and challenges that could negatively impact the perception of them. AI tools may, for example, be able to enhance election administration efficiency, improve voter information access, and assist in detecting potential security threats; at the same time, they can be weaponized to generate and spread election-related misinformation at unprecedented scale and sophistication. The State Board of Elections (SBE) takes this seriously and has, for example, created an election misinformation reporting portal and implemented active automated identification and take-down of election misinformation on media. Ongoing work here will become more critical as AI-generated content becomes increasingly difficult to distinguish from authentic communications, potentially impacting voter trust and participation.

• ​​​ Plan:
• ​Meet with elections researchers, civil society experts, other state election officials, and other stakeholders to identify actionable approaches and policies to leveraging AI to improve election operations and/or combat the impact of deepfakes, disinformation, and misinformation in future elections.
• Evaluate if enhancements to SBE’s election mis-, dis-, and mal-information reporting portal are needed to better monitor, collect, report on, and act on instances of election-related activity in this area.
• Stakeholders:
• ​​​ Owners:
• SBE
• Involve​:​
• GOHS
• OAG -Elections Infrastructure Information Sharing and Analysis Center (E or I-ISAC)
• Relevant researchers in academia / civil society / think tanks
• Technology platform providers
• Projected Timeline:
  
• ​​​ By July 2025: Complete the initial phase of portal data collection
• By Dec. 2025: Complete initial portal data analysis and reporting​

​​​​Procurement

​​​​Relevant SB818 requirement:

​​​​“A plan to study methods to ensure that there is diversity in contract awards and training programs related to artificial intelligence in the State, including racial diversity.”

​​​​“A plan to study the procurement of systems that employ artificial intelligence, including efforts to increase competition and assurance that contracts retain sufficient data privacy protection against vendor lock–in.”

​​​​ The procurement process is a crucial lever for operationalizing AI governance and adoption in the State, given much of agencies’ use of AI will be via acquisition. The domains of AI and procurement intersect in a variety of ways, and we will explore and adopt best practices, leveraging existing mechanisms as possible.

• ​​​ Plan:
• Literature review and discussions with exemplar states, federal agencies, AI governance researchers, and other relevant parties. Focus on mapping different approaches to procuring AI solutions that can power MD’s AI Enablement strategy; operationalizing AI governance via procurement; using AI to improve the procurement process; and adopting a “buyers’ guide” to help procuring teams make better decisions around build vs. buy, product selection, total cost of ownership, and other considerations.
• Leverage inputs to inform ongoing efforts to create a policy, best practices, and guidance for procuring AI systems and evaluate inserting additional RFP/contract requirements pertaining to data privacy, security, and other areas pertinent to operationalizing the State’s AI principles (including as part of the Data Use Agreement). Integrate effectively with the broader efforts to operationalize robust AI governance in the State.
• Leverage Senior Procurement Advisory Group (SPAG) meetings to provide training to procurement officers as needed, particularly around methods available for completing Procurement Review Group (PRG) worksheets for establishing socioeconomic goals on contracts.
• Stakeholders:
• ​​​ Owners:
• DGS
• DoIT
• Involve​:​
• Governor’s Office of Small Business (MBEs)
• Projected Timeline:
• Throughout the year.​

2026 and Beyond

​​​​Given the rate of change, predicting more than a year out is difficult. We are taking an iterative approach to most of the workstreams, which allows us to test and learn more quickly. The outputs from that test, learn, and iterate process, as well as the 2025 studies in key domains, will allow us to set future strategies rooted in real needs rather than theory. Assuming we are successful in executing our 2025 strategy, a number of themes are likely to increase in importance in 2026:

• Ensuring the entire State workforce has negotiated, cost-effective access to modern, empowering AI tools in key, validated productivity categories to remove time, cost, friction, and drudgery from their workflows.
• Improving our capacity to move successful PoCs and experiments to deployed, scaled systems “in production,” with robust monitoring, to power improvements in key domains while avoiding the fate of dozens of PoCs taking resources but not yielding benefits.
• Shifting more focus to the underlying data - further increasing maturity in governance, management, security, privacy, labeling, and “readiness” - given its centrality and complexity in leveraging AI tools.
• Developing the capability to responsibly build on open source models in addition to leveraging cutting edge closed models, with clear “build vs. buy” pathways.
• Operationalizing more mature operations frameworks (e.g. - “AIOps”) as more State services adopt AI-powered components.
• Ensuring we have the right frameworks and infrastructure to experiment with, adopt, and govern even more impactful near term opportunities like agentic systems.
• Reducing silos between the State, academia, industry, startups, civil society, and other government entities in MD, to encourage a more robust, collaborative, and productive AI ecosystem.
• Maturing the AI Enablement Team into a robust incubation shop to help the State adopt emerging technologies. Overall, success means adopting AI in ways that moves the needle on the Governor’s priority areas, in a manner that fully takes advantage of thewealth of intellectual, institutional, and people-powered resources across Maryland.

​​​​​ Overall, success means adopting AI in ways that moves the needle on the Governor’s priority areas, in a manner that fully takes advantage of the wealth of intellectual, institutional, and people-powered resources across Maryland.