DC's AI Values
AI promises to improve the delivery of government services in every jurisdiction. And while its incorporation into the work of the District government is essential, the deployment of powerful tools can bring about unintended and harmful impacts if risks are not carefully weighed in advance, and possible mitigation strategies carefully considered. Accordingly, Mayor's Order 2024-028 requires any agency planning to deploy an AI tool first to verify whether such deployment is in alignment with the following AI Values, assess impacts to these AI Values, consider what controls might be used to mitigate negative impacts, and document its review:
Clear Benefit to the People
No AI tool shall be deployed unless the deploying agency has verified that the deployment will benefit District residents. An agency deploying an AI tool shall articulate its purpose in deploying the tool, who will benefit from the planned deployment, and what alternative means of achieving the purpose and the benefit were considered. An agency’s verification that deployment of an AI tool benefits District residents might include, for example, whether such deployment allows the agency to deliver more services, deliver services more quickly, deliver services with greater accuracy, or deliver services of greater quality. Agencies may analyze potential benefits from a broad perspective, for example identifying benefits that might be conferred only upon a limited population, benefits that may not mature immediately but which may be realized in the future, and benefits which are shared in common outside the District such as positive impacts on the environment. In exercising such breadth of scope, agencies should at the same time be careful to evaluate critically whether an apparent benefit clearly flows from the deployment and whether such benefits are tenuous, speculative, or inconsequential.
Safety and Equity
No AI tool shall be deployed unless the deploying agency has carefully considered how deployment will impact safety and equity in the District, and weighed such impacts against the impacts of alternative courses of action, including maintenance of the status quo. An agency deploying an AI tool shall clearly explain its consideration of any of the following risks entailed by the deployment: risk of direct and indirect physical harm (including exacerbation of existing harm), risk of deprivation of rights, and risk of exacerbating inequity.
No AI tool shall be deployed unless the deploying agency has articulated how deployment will maintain meaningful accountability for government action. An agency deploying an AI tool shall explain what steps it has taken to preserve human accountability for every action or determination flowing from the AI tool. An agency deploying an AI tool shall also articulate its plan for pre- deployment and continuing testing and validation of the AI tool’s performance in the environment in which it is deployed.
No AI tool shall be deployed unless the deploying agency has articulated how the deployment will be conducted transparently. An agency deploying an AI tool shall explain when and how it will disclose when persons are directly interacting with an AI tool, like an AI-powered chat bot, and not a human. An agency deploying an AI tool should also explain what steps it has taken to avoid the release of AI-generated content without substantive review and approval by a human, in contexts where such unsupervised release would not be appropriate.
No AI tool shall be deployed unless the deploying agency has considered the deployment’s sustainability. An agency deploying an AI tool shall explain what steps it has taken to assess the environmental impact of both use and development of the relevant AI tool, and the impact deployment will have upon employee displacement and job quality of District employees. An agency deploying an AI tool shall explain how it has accounted for the risk of future unilateral price increases by the provider of the AI tool within the context of market concentration and sunk costs.
Privacy and Cybersecurity
No AI tool shall be deployed unless the deploying agency has carefully considered the deployment’s impact upon privacy and cybersecurity. An agency deploying an AI tool shall explain how it plans to ensure the deployment does not violate privacy rights or undermine the privacy of residents, employees, and visitors.
Advisory Group on AI Values Alignment
Mayor Bowser has established an Advisory Group on Artificial Intelligence Values Alignment to to engage with internal and external stakeholders on the proper intersection of the District’s AI Values and emerging and future AI technologies to ensure all stakeholder perspectives are heard, and to advise the Mayor on the refinement and alignment of the District government’s AI Values on an ongoing basis. Interested in being a member of the Advisory Group? Apply today.
The Advisory Group will:
- Hold public listening sessions to solicit public input on the appropriate government use of AI tools.
- Upon request of the Mayor, review a specific AI tool, deployment, or general AI technology and advise the Mayor on how the tool, deployment, or technology might best be brought into alignment with the District’s AI Values.
The Advisory Group shall comprise the following members, all of whom shall be appointed by, and serve at the pleasure of, the Mayor:
The following public members, who shall not be employees of the District government:
- One public member to serve as the Community Co-Chair of the Advisory Group;
- Additional public members whose expertise shall include: racial equity; disability rights; labor relations; criminal justice; education; healthcare; privacy; cybersecurity; and transparency/open government.
Public members of the Advisory Group shall serve at the pleasure of the Mayor from the date of their appointment until December 31, 2026.
AI Strategic Benchmarks
- By May 8, 2024, the Office of the Chief Technology Officer shall submit a standard privacy review processes governing: the approval of deployments of AI tools which will make use of enterprise data classified higher than Level 0; and the regular review of the District government’s Enterprise Data Inventory in light of privacy risks arising from emerging or future AI functionalities.
- By May 8, 2024, the Office of the Chief Technology Officer shall submit standard cybersecurity review processes governing: the approval of deployments of AI tools which interact with the District government’s information technology environment, including deployments that precede the effective date of this Order; and the regular review and assessment of emerging, AI-powered cybersecurity threats.
- By August 8, 2024, the Department of Human Resources and the Department of Employment Services shall submit an integrated recruitment and workforce development plan. The plan shall identify existing recruitment and workforce development resources relevant to the emerging AI job sector and shall articulate strategies for growing the AI-ready workforce in the District of Columbia and recruiting AI talent into the District government workforce.
- By August 8, 2024, the Department of Human Resources and the Office of the Chief Technology Officer, in consultation with the Office of Labor Relations and Collective Bargaining, shall submit a body of comprehensive training materials on the proper use of AI tools by District government personnel.
- By September 6, 2024, the Office of Contracting and Procurement shall submit a mandatory AI procurement handbook. This AI procurement handbook shall provide guidance to agencies in the basic capabilities of broad categories of AI tools, structuring and scoping AI tool procurements, and effectively monitoring performance of procured AI tools.
- Before October 1, 2024, a first cohort of agencies selected by the City Administrator, before October 1, 2025, a second cohort of agencies selected by the City Administrator, and before October 1, 2026, the final cohort of agencies selected by the City Administrator, shall submit their agency-specific AI strategic plans. These agency- specific AI strategic plans shall identify: aspects of the agency’s mission where currently available AI tools may improve agency performance, areas where the incorporation of AI tools may present new risks, safeguards the agency will utilize to mitigate risks presented by the deployment of AI tools, and the agency’s AI tool deployment stakeholder engagement plan.
The Mayor will establish within in the Executive Branch of the Government of the District of Columbia the Artificial Intelligence Taskforce or AI Task Force.
The purpose of the AI Taskforce is to provide deep AI strategy and governance expertise and support to the Mayor and agency leadership, and to coordinate agency efforts to meet the AI Strategic Benchmarks.
The AI Taskforce shall:
- Upon request of the Mayor or her designee, review and provide technical advice on any specific AI tool, deployment, or general AI technology;
- Provide regular reports to the City Administrator concerning agency progress in meeting the AI Strategic Benchmarks;
- Furnish AI expertise and guidance as necessary to ensure agencies meet the AI Strategic Benchmarks;
- Develop a form whereby agencies can record their assessment of each AI tool deployment’s alignment with the District’s AI Values;
- Undertake such other functions as may be requested by the Mayor.
The AI Taskforce shall comprise five members, all of whom shall be District government employees,one each with expertise in:
- AI and technology law;
- AI and technology enablement;
- AI and data science;
- AI and software development; and
- AI and cybersecurity.
One member of the AI Taskforce shall be designated by the Mayor as Taskforce Lead.
Members of the AI Taskforce shall be appointed by the Mayor and shall serve at the pleasure of the Mayor from the date of their appointment until December 31, 2026, unless earlier removed by the Mayor.
The AI Taskforce shall be composed of District government full-time employees, serving the District government within the ordinary course of their employment. The AI Taskforce is not a public body, and does not conduct meetings, as such terms are defined by the District of Columbia Open Meetings Act, effective March 31, 2011 (D.C. Law 18-350; D.C. Official Code §§ 2-571 et seq. (2016 Repl.)).
The Office of the Chief Technology Officer shall provide administrative support to the AI Taskforce.
The members of the AI Taskforce shall serve without compensation.
The AI Taskforce shall sunset on December 31, 2026.