The Complete Guide to Using AI in the Government Industry in Washington in 2025

Washington, DC newcomers to public service should pay attention: federal guidance now treats AI as a practical toolkit for improving mission delivery, not a mysterious tech buzzword - see the GSA AI Guide for Government (printable) for a clear, evolving playbook on organization, responsible deployment, and workforce strategy (GSA AI Guide for Government (printable)).

Practical primers like the Code for America AI in Government cheat sheet decode terms (LLMs, entity resolution, supervised learning) and show how simple pilots - think using entity resolution to untangle messy records - can free staff for higher‑value work (Code for America AI in Government cheat sheet).

For beginners who want hands‑on skills, a short, applied course like Nucamp's AI Essentials for Work registration teaches prompt writing, tool use, and real workplace projects so DC teams can move from cautious curiosity to responsible experimentation without needing a PhD (Nucamp AI Essentials for Work registration).

Washington, D.C.'s policy moves the conversation from “can we use AI?” to “how should we use it?” by requiring agencies to verify alignment with six clear AI Values before any deployment: Clear Benefit to the People; Safety & Equity; Accountability; Transparency; Sustainability; and Privacy & Cybersecurity - details and reporting tools live in Mayor's Order 2024‑028 and the District's AI Values and Strategic Plan (Mayor's Order 2024‑028 and AI Values).

The Order doesn't leave agencies to guess: it mandates documented alignment reviews, agency‑specific AI strategic plans, privacy/cybersecurity review processes and an AI procurement handbook on fixed timelines, and it creates two governance bodies - a public Advisory Group on AI Values Alignment and an internal AI Taskforce to vet technical questions and coordinate benchmarks through 2026.

Practical safeguards are baked in too: label AI‑generated content, preserve human accountability, and test tools in controlled environments (the strategy even includes a SCIF “testing sandbox” for secure experimentation) (DC's AI testing sandbox and SCIF), so agencies must show measurable benefits to residents, not speculative promises, before a single model goes live.

“We are going to make sure DC is at the forefront of the work to use AI to deliver city services that are responsive, efficient, and proactive. With these guiding values, we will make sure that when we use AI, we are responsible and use it in a way that aligns with our DC values.” - Mayor Bowser

In 2025 the federal landscape is moving fast and will directly shape how Washington, D.C. agencies adopt AI: Executive Order 14179 set the tempo for a deregulatory, innovation-first playbook and OMB's implementing memoranda (M‑25‑21 and M‑25‑22) turn that playbook into concrete steps - expect Chief AI Officers, formal governance boards, mandatory procurement guardrails, tighter data‑rights clauses, and explicit guidance to avoid vendor lock‑in when buying AI (OMB memos M‑25‑21 and M‑25‑22 implementing Executive Order 14179).

The White House's July AI Action Plan and three linked Executive Orders amplify three national priorities - accelerate AI innovation, build secure AI infrastructure (data centers and permitting), and lean into international AI diplomacy - while adding procurement rules that will ripple through the vendor market (White House AI Action Plan and Executive Orders overview).

Agencies using “high‑impact” AI face minimum risk‑management practices (pre‑deployment testing, documented impact assessments, ongoing monitoring, human oversight) and deadlines for governance and reporting, so District teams should map local AI Values and Mayor's Order requirements to these federal timelines and contract terms to avoid surprises when procuring or operating AI (Federal memorandum on requirements for high‑impact AI used by agencies); think of it as treating AI projects like mission‑critical IT upgrades - CAIOs, sandboxes, and testbeds are now part of the job.

DEI described as “one of the most pervasive and destructive” ideologies in AI contexts

Congress did not produce a single, comprehensive AI Act in 2025, but several competing blueprints and the White House's pro‑innovation push have shaped what a federal law might look like - and why the District should watch closely.

Lawmakers circulated risk‑based proposals (the SAFE Innovation Framework, a Bipartisan U.S. AI Act draft, and a National AI Commission proposal) that range from transparency and licensing regimes to a congressional commission charged with crafting a national risk framework, while the White House's AI Action Plan presses a deregulatory, infrastructure‑first agenda and even signals withholding discretionary funds from states with burdensome AI rules (a move that could affect local initiatives) (see the University of Chicago comparison of U.S. proposals to the EU AI Act and the Jenner client alert on state AI regulation).

With Congress divided and federal bills slow to land, states have filed a flood of local measures - more than a thousand AI bills in early 2025 - so for DC officials the immediate questions are practical: which federal guardrails (if any) will preempt city rules, how procurement and funding guidance from the White House will affect DC grants, and whether a future U.S. AI law will mirror the EU's risk‑based model or favor the U.S. emphasis on innovation and national security.

Washington, D.C. agencies should anchor AI programs in the National Institute of Standards and Technology's work: NIST's AI Risk Management Framework (updated as AI RMF 2.0) has become the go‑to playbook for mapping, measuring, managing and governing risks across the AI lifecycle, and companion releases in 2024 give practical, testable steps for generative models and secure development.

Key NIST outputs - the Generative AI Profile and secure software development practices (SP 800‑218A), plus tooling like the Dioptra security testbed and the ARIA model‑evaluation pilot - turn abstract principles into activities (red‑teaming, field testing, provenance and measurement) that agencies can adopt before buying or deploying systems; see WilmerHale's summary of the July 26, 2024 releases for the testbed and guidelines (WilmerHale summary of NIST risk mitigation guidelines and Dioptra testbed) and Diligent's plain‑language guide to why AI RMF 2.0 is effectively a government‑aligned “gold standard” for practical governance (Diligent guide to the NIST AI Risk Management Framework).

The guidance is voluntary but influential - treat it as the checklist that makes pilots auditable, vendors accountable, and red‑team exercises feel less like theory and more like infrastructure protection.

Buying AI in the District now starts long before a purchase order: Mayor's Order 2024‑028 forces agencies to prove an AI project's “clear benefit to the people,” complete privacy and cybersecurity reviews, and follow OCTO's procurement playbook (including the mandatory AI procurement handbook and agency alignment reports) before a tool is deployed (Mayor's Order 2024‑028: DC AI Values and Strategic Plan).

Practically, that means solicitations should demand documented alignment with DC's values, clear data‑use rules, human‑in‑the‑loop obligations, and exit or “decommissioning” terms to guard against surprise price hikes or vendor lock‑in - federal procurement trends are already pressing for expanded government rights over models, datasets, and localization requirements, so city contracts must be drafted with those pressures in mind (Government contracting and AI trends 2025: rights, datasets, and localization).

Procurement teams should modernize workflows (digital procurement platforms, automated bid evaluation, supplier‑diversity and sustainability scoring) and bake in continuous monitoring and red‑team testing as conditions of award; DC's pilot with MIT GOV/LAB and Stanford DEL for deliberation.io shows how procurement can pair careful public testing with community input before scaling (OCTO deliberation.io pilot with MIT GOV/LAB and Stanford DEL and public listening plan).

The upshot: treat AI buys like mission‑critical IT upgrades - insist on auditable data rights, privacy controls, cybersecurity attestations, and contract terms that protect residents if a vendor changes price, policy, or access.

Intellectual‑property and legal risk management are now core parts of any DC agency AI plan: the U.S. Patent and Trademark Office has sharpened subject‑matter eligibility rules for AI inventions (effective July 17, 2024) and separate inventorship guidance that explains when a human contributor is required for a patent, so teams drafting procurements or collaborating with innovators should demand clear inventor attribution and claims that show concrete technical improvements rather than abstract data processing (USPTO AI subject‑matter eligibility guidance (July 17, 2024) and related inventorship guidance).

Practitioners must also treat AI‑assisted filings and vendor tools with the same duty of candor, signature verification, and confidentiality that apply to any legal submission - oversharing training data or client files with third‑party models can trigger privacy, export‑control, or even national‑security complications, so procurement language should lock down data flows and audit rights (USPTO guidance on using AI-based tools for legal filings and procurement).

On copyright, the Copyright Office's multipart study (Parts 1–3, with Part 2 on output copyrightability and a pre‑publication Part 3 on training) maps how generative outputs and training sets are being evaluated - an essential read for DC teams deciding what materials can be used to train models or published as agency content (U.S. Copyright Office study: Copyright and Artificial Intelligence).

Bottom line for District leaders: bake inventorship, patent‑eligibility tests, confidentiality safeguards, and clear training‑data rules into contracts up front - even one stray upload to a public chatbot can undo privacy protections and procurement gains.

“The USPTO remains committed to fostering and protecting innovation in critical and emerging technologies, including AI,” said Kathi Vidal, Under Secretary of Commerce for Intellectual Property and Director of the USPTO.

Washington, D.C. is treating AI workforce development as an operational priority, not an afterthought: Mayor's Order 2024‑028 directed the Department of Human Resources and the Department of Employment Services to deliver an integrated recruitment and workforce development plan and comprehensive staff training materials by August 8, 2024, while OCTO and the Advisory Group are running public listening sessions and trainings to connect residents and practitioners to policy and practice (see the DC AI Values and Strategic Plan DC AI Values and Strategic Plan).

City efforts are paired with practical learning pathways - local pilots like the DC Public Library's free AI Upskilling Cohort (cohorts of five to seven adults, workshops and hands‑on projects at the Martin Luther King Jr.

Memorial Library) give residents portfolio‑building, career‑relevant experience this summer, and tailored options such as the Public Health Foundation's AI workshops help agencies embed safe, problem‑solving uses of AI into everyday public‑service work (DC Public Library AI Upskilling Cohort details and enrollment, Public Health Foundation AI technical assistance and training information).

The result: small, practical learning loops - think a five‑person cohort that finishes a real project and hands a vetted prompt library to an agency team - help retain staff, reduce fear of automation, and make the “so what?” tangible: faster, audited services that still keep humans in the loop.

“AI literacy is the next essential skill people need to succeed in today's workforce, and this cohort delivers training in a way that works for busy adults.” - Chelsea Kirkland, DC Public Library

Washington, D.C. stakeholders should plan for an industry shakeup driven by huge private investment in AI infrastructure: the Stargate Project - announced as a new company led by SoftBank and OpenAI with Oracle and MGX among initial backers - promises up to $500 billion over four years (with $100 billion to start) to build U.S. AI data centers and related power and compute capacity, beginning with a campus in Abilene, Texas (Stargate Project official press release).

That scale matters for the District because it will reverberate through procurement, workforce planning, and energy siting - expect tougher competition for GPUs, skilled construction and operations talent, and sharper scrutiny on grid impacts and sustainability as analysts flag supply‑chain and power constraints that could crowd out other projects (DatacenterDynamics analysis of Stargate project risks and industry implications).

Regulators and agency buyers in DC should watch near‑term capacity numbers (OpenAI and partners are targeting multi‑gigawatt builds, with reports of 5+ GW under development and a 10 GW goal), prepare contract language that guards against vendor lock‑in, and link workforce training to real construction and operations jobs - if Stargate or similar builds scale, the “so what” is concrete: entire regional grids and vendor markets will feel the pull of AI demand, and city teams that pair procurement discipline with targeted upskilling will capture both the benefits and the jobs.

“This infrastructure will secure American leadership in AI, create hundreds of thousands of American jobs, and generate massive economic benefit for the entire world.” - OpenAI / Stargate announcement

Getting started with AI in the District means pairing bold curiosity with the guardrails DC has already written into law: map every pilot to DC's AI Values and Strategic Plan so projects demonstrate a clear benefit to residents, preserve human accountability, and pass OCTO's privacy and cybersecurity reviews (DC AI Values and Strategic Plan).

Use the Mayor's Order milestones and the newly formed AI Taskforce as a roadmap - start with small, accountable pilots that deliver measurable service improvements, involve the Advisory Group's public listening sessions, and treat procurement like a mission‑critical IT upgrade so contracts lock down data rights and decommissioning terms (see the Mayor's press release outlining these actions) (Mayor's press release on DC's AI Values).

For teams and individual staffers who need practical skills now, short applied learning - such as a 15‑week AI Essentials for Work course - turns policy into practice: think a five‑person cohort that finishes a real project and hands an audited prompt library to an agency team, lowering fear of automation and producing faster, more equitable services; explore options and register to build those real‑world skills (Register for Nucamp AI Essentials for Work).

“We are going to make sure DC is at the forefront of the work to use AI to deliver city services that are responsive, efficient, and proactive. With these guiding values, we will make sure that when we use AI, we are responsible and use it in a way that aligns with our DC values.” - Mayor Bowser