Day 2 of 5
Day 2 AI for Federal Employees / Day 2

AI Tools Approved for Government Use

FedRAMP authorized AI tools, data classification levels, and what you can use today — without procurement. By the end of this lesson you will know exactly which tools are available to you and at what data sensitivity level.

45 min read 1 exercise Reference tables included

Why Tool Selection Is a Security Question, Not Just a Preference

In the private sector, picking an AI tool is mostly a question of cost and features. In the federal government, it is primarily a security and compliance question. Using the wrong tool with the wrong data can expose sensitive information, violate federal data handling requirements, or create legal liability for you and your agency.

The governing framework for federal cloud tools — including AI — is FedRAMP: the Federal Risk and Authorization Management Program. FedRAMP provides a standardized approach to security assessment for cloud products and services used by the federal government. If a tool is not FedRAMP authorized for the data level you need, you should not use it with that data. Full stop.

This is not bureaucratic restriction for its own sake. It is a reasonable professional standard. Consumer AI tools route your inputs through commercial cloud infrastructure that has not been authorized for federal data. The risk is real.

Understanding Data Classification Levels

Before you can know which tools are appropriate, you need to understand what kind of data you work with. Federal data is classified along the DoD Impact Level (IL) framework and the broader Federal Information Processing Standards (FIPS) framework. Here is the practical breakdown:

Level What It Covers Examples AI Implication
Unclassified (Public) Information that is publicly available or has no harm if released Published reports, public-facing documents, general policy information Consumer tools (free tier) are appropriate for drafting and research
CUI (Controlled Unclassified) Sensitive but unclassified — requires protection under law or regulation PII, FOUO documents, pre-decisional policy, acquisition-sensitive data Requires FedRAMP Moderate or higher authorized tools
IL4 Controlled Unclassified / sensitive national security data DoD Controlled Unclassified Information, some mission-sensitive data Requires FedRAMP High or DoD IL4 authorized tools
IL5 National Security Systems / higher sensitivity CUI Higher-sensitivity DoD data, some classified national security information Requires specific IL5 authorized tools — very limited AI options
IL6 (Secret) Classified / Secret Classified national security information Only DoD-authorized classified AI environments — extremely limited
Key rule: Never input CUI, PII, acquisition-sensitive, or pre-decisional policy content into a consumer AI tool. This includes the free tiers of ChatGPT, Claude.ai, and Google Gemini. These tools are appropriate for working with publicly available information only.

FedRAMP Authorized AI Tools (As of April 2026)

The FedRAMP marketplace has grown rapidly. Here are the most commonly deployed AI tools in federal agencies, organized by capability:

Large Language Models / AI Assistants

Microsoft Azure OpenAI Service
Microsoft / OpenAI — FedRAMP High Authorized
Provides GPT-4 class models (including GPT-4o) through Azure Government cloud. Most widely deployed AI service in federal agencies. Available through agency Microsoft enterprise agreements.
IL4 Available IL5 (DoD approval)
Claude on AWS GovCloud (Amazon Bedrock Government)
Anthropic / Amazon — FedRAMP High Authorized
Claude models (Claude 3 Sonnet, Claude 3 Opus) available through AWS GovCloud via Amazon Bedrock. Appropriate for CUI and sensitive federal data when deployed through agency AWS GovCloud accounts.
IL4 Available
Google Vertex AI (Google Cloud Government)
Google — FedRAMP High Authorized
Gemini models and Google's AI suite available through Google Cloud for Government. Less widely deployed than Azure OpenAI in civilian agencies, but growing in DoD contexts.
IL4 Available

AI-Enhanced Productivity Tools

Microsoft Copilot for M365 Government
Microsoft — FedRAMP High Authorized
AI assistant integrated into Word, Excel, PowerPoint, Outlook, and Teams for government. Available to agencies with M365 Government licensing. One of the fastest-growing AI deployments in civilian agencies.
IL4 (GCC High) Unclassified
AWS AI Services (Comprehend, Textract, Rekognition)
Amazon — FedRAMP High Authorized
Document processing, natural language analysis, and image/document recognition services. Widely used for automating document-heavy workflows in federal agencies.
IL4 Available

Free Tools You Can Use Today (for Non-Sensitive Work)

For unclassified, non-sensitive, publicly available information — the kind of work that does not involve PII, CUI, or pre-decisional content — you can use consumer AI tools right now, at no cost, without procurement:

Tool Free Tier Best For Data Rule
Claude.ai Claude 3.5 Haiku — generous free tier Long document summarization, drafting, structured analysis Public/unclassified only. No PII. No CUI.
ChatGPT (OpenAI) GPT-4o mini — free with account Research, brainstorming, first-draft writing Public/unclassified only. No PII. No CUI.
Google Gemini Gemini 1.5 Flash — free tier Research, summarization, document analysis Public/unclassified only. No PII. No CUI.
Microsoft Copilot (free) Copilot.microsoft.com — free Web research, drafting, summarization Public/unclassified only. No PII. No CUI.
Start here: If your agency has not yet deployed an enterprise AI solution, you can begin building your personal AI proficiency using these free tools on unclassified tasks. Draft public-facing communications. Summarize published reports. Research policy questions using public sources. This builds the muscle before the enterprise tools arrive.
Day 2 Exercise

Try One AI Tool on a Real (Non-Sensitive) Work Task

Use one of the free tools above to complete a real task from your work — something involving publicly available information only. Here are proven starter tasks:

  • Summarize a published report or policy document. Paste in the public text and ask: "Summarize the key findings and recommendations in 5 bullet points for a senior executive audience."
  • Draft talking points for a meeting. Describe the meeting context (no sensitive details) and ask for a structured talking points outline.
  • Research a policy question. Ask the AI to explain a piece of legislation, regulation, or policy area in plain English.
  • Improve a public-facing document. Paste in a draft of publicly releasable text and ask for editing suggestions that improve clarity.

After your attempt, answer these three questions in writing:

  1. What did the AI do well?
  2. What did it get wrong or miss?
  3. What would I need to verify before using this output?

This reflection practice — what worked, what didn't, what needs verification — is the habit that separates effective AI users from people who get burned by AI mistakes.

Key Takeaways from Day 2

Need enterprise AI for your agency?

Our federal consulting practice helps agencies navigate FedRAMP procurement, deploy authorized AI tools, and train their workforce. Contact us to discuss your agency's needs.

Start the Conversation →