“Artificial intelligence” (AI), as used here, refers to a machine-based system that can make predictions, recommendations, or decisions. When used responsibly, it is a powerful technology that can improve federal government operations and service delivery. Within the broad field of AI, Generative AI (GenAI) is a rapidly evolving technology that can generate content, including text, audio, images, or video, in response to a prompt or input data. It includes conversational AI applications (for example, OpenAI ChatGPT, Google Gemini, Meta LLaMa, Anthropic Claude) and image generation AI applications (for example, OpenAI DALL-E, Midjourney, Stability AI Stable Diffusion, Adobe Firefly), as well as applications to generate speech, video, or programming code (for example, GitHub Copilot, Amazon CodeWhisperer). GenAI technology is trained on large amounts of data, which can include publicly available text, images, audio, and videos scraped from the Internet.

What does this mean to me as a federal employee?

GenAI has the potential to improve the way the federal workforce delivers results for the public. Federal employees can leverage GenAI to enhance creativity, efficiency, and productivity. Federal agencies and employees are encouraged to consider how best to use these tools to fulfill their missions.

As a federal employee, you should carefully consider the benefits and risks of using GenAI in your work and use these tools in a safe, secure, and responsible way.

How should I use this webpage?

• Learn about GenAI’s potential benefits and risks.
• Explore best practices for safely, securely, and responsibly using GenAI in your work.
• Find ways to learn more about GenAI. Note that this webpage is not comprehensive. You should review and follow your agency’s relevant policies and procedures.

This content may be periodically updated to account for advancements in GenAI and evolving workforce needs.

Potential benefits and risks of using GenAI

Potential benefits

GenAI applications offer the federal workforce many potential benefits related to productivity and agency operations, including:

• Increasing the efficiency of manual or repetitive tasks;
• Synthesizing information, such as summarizing meeting notes or transcripts, or extracting the main idea or key elements from documents;
• Brainstorming and ideating;
• Drafting and designing communication products, including generating outlines or text and images for reports, presentations, emails, web, or social media;
• Improving and refining writing, including providing grammar, spelling and style suggestions and transforming complex concepts into plain language;
• Translating products into other languages;
• Enhancing customer experience by creating customized responses to inquiries;
• Improving accessibility through features such as audio and video transcription;
• Accelerating software development by generating, explaining, or debugging code, or by translating code from one programming language to another;
• Analyzing and deriving insights from datasets and generating graphs.

Potential risks

The use of GenAI also presents potential risks that need to be managed so that the federal workforce can leverage this technology responsibly, including:

• Fabrications and errors, such as reasoning mistakes or factually inaccurate text, speech, images, or videos;
• Plagiarism or copyright violations;
• Security vulnerabilities of the tools;
• Privacy risks, including from unintended disclosure of training data, prompts, or input data;
• Amplification of biased, inequitable, or discriminatory content;
• Legal or ethical risks, such as AI-generated content that encourages illegal or unethical behavior;
• Risks to agency credibility if content is inaccurate, unreliable, or discriminatory;
• Uneven performance depending on languages, fields, and applications;
• Harmful use by individuals external to the federal government to create misinformation, impersonate elected officials, or spam an agency.

Recommendations for federal employees using GenAI

When using GenAI in the workplace to conduct official business, remember to:

• Follow your agency’s policies and procedures, including for AI governance, data governance, privacy, and security.
• Use your agency’s GenAI resources to assist with your official duties and not for personal use. Expect that your use of GenAI technology may be logged and monitored.
• Consult with your agency for a list of approved GenAI technologies that may be used for official business. Follow agency directives on when and how to use GenAI technology through public interfaces (that is, any applications or software that use GenAI that are available to the general public) or through your agency’s secure enterprise system as there are risks associated with either type of access, including risks related to storage, reuse, or disclosure of information.
• Understand any restrictions on information that can be input into GenAI interfaces and only input information that is allowed. Do not input non-public information into public GenAI interfaces.
• Oversee the use of GenAI technology. Do not leave GenAI technology to operate autonomously without human accountability and control.
• Review all AI-generated materials, including sources cited, to check that they are valid, accurate, complete, and safe.
• Review AI-generated material for potentially biased, harmful, stereotypical, or otherwise offensive language or images. Do not tamper with or attempt to turn off guardrails against unsafe, offensive, or misleading inputs or outputs of GenAI systems.
• Maintain integrity and avoid ethical misconduct (for example, fabrication, falsification, or plagiarism), including in any AI-generated material.
• Follow your agency’s policy on when and how to disclose uses of GenAI (for example, labels, watermarks, or disclaimers on AI-generated material).
• Use GenAI tools only for appropriate purposes and not to generate malicious, inappropriate, or illegal material, such as malware, deepfakes, toxic content, or other material considered not suitable for work.

Examples of GenAI uses

Some federal agencies already use AI to improve operational efficiency and effectiveness. You can find your agency’s reported federal AI use cases on AI.gov. This inventory will be updated with GenAI use cases as agencies continue to explore adopting this technology.

Below are examples of potential uses of GenAI that you may encounter as part of your work, which provide additional, use-specific considerations and recommendations. These do not present a comprehensive set of possible uses or recommendations that federal employees should abide by when using GenAI. Work with your supervisor and relevant agency officials to understand your agency’s particular policies, including whether GenAI is appropriate for your use case, and which GenAI technology should be used.

Using GenAI to summarize or analyze information

Use: GenAI technology can be used to synthesize information, such as summarizing meeting notes and transcripts, extracting main ideas from documents (like research papers or policies), or analyzing data (for example, producing summary statistics or graphs). It also can answer questions about a particular topic appearing in documents or data.

Considerations: GenAI technology may omit essential details from summaries, fabricate information that is not included in the documents or data, or include additional materials from its training set that are not in the source inputs. GenAI technology may also introduce biased or outdated language when synthesizing. It can also be subject to reasoning and mathematical errors.

Recommendations: When using GenAI to summarize information or analyze data, you should:

• Review the accuracy and validity of AI-generated summaries to detect and correct potential issues, including factual inaccuracies, logical reasoning or mathematical errors, and inappropriate language.
• Verify that the AI’s summary is consistent with the inputs, is complete, and does not include information beyond the scope of the documents or data.
• Check that AI-generated content has appropriate citations, references, or other attributions for information used.

Using GenAI to create communications materials

Use: GenAI can be a powerful tool for developing communications for various reading levels, languages, platforms, and audiences. GenAI technology can be used to draft content, develop outlines, improve grammar and spelling, transform content into plain language, translate materials in multiple languages, or generate audio or visual assets.

Considerations: Because GenAI technology’s training data may not equally represent non-English languages and perspectives from underserved communities, generated communication outputs may be inaccurate, biased, or harmful. Training data may also be outdated, so AI-generated material may not include recent events or perspectives.

Additionally, the use of GenAI technology to produce communications may not be appropriate for sensitive topics or events.

Recommendations: When using GenAI to produce communications materials, you should:

• Review and fact-check the content produced by GenAI tools with subject matter experts. For example, if translating products to another language or for a new audience, your agency may require that persons proficient in that language and with knowledge of that cultural context review these materials.
• Select AI-generated text and images that are respectful and inclusive and review final products for accessibility.
• Check if AI-generated content includes copyrighted material and consult with your agency legal counsel about use of such materials.
• Verify your agency’s requirements on disclosing use of GenAI for materials produced for external and internal audiences.

Using GenAI for software and code development

Use: GenAI technology has the potential to improve the efficiency and speed of software and code development. For example, it may help with programming code generation, review, explanation, and debugging.

Considerations: AI-generated code may include bugs or introduce security vulnerabilities. It can also introduce new libraries and dependencies, including nonexistent or malicious packages. It may also borrow from open-source code without attribution.

Recommendations: When using GenAI for software or code development, you should:

• Verify that you have the necessary agency approvals for conducting code generation and review with GenAI, prior to leveraging GenAI for this purpose. Review your agency’s policies and industry best practices for generating software and code.
• Conduct review and testing of AI-generated code before deploying it. Fix bugs, address vulnerabilities, and check the legitimacy of AI-generated libraries and dependencies. Verify that AI-generated code is performing as intended.
• Familiarize yourself with your agency’s requirements for code documentation and storage and for disclosure of use of GenAI. Pay special attention to restricted activities, such as prohibitions on inputting internal code (for example, code from an agency’s enterprise code repository that is not publicly available) into public interfaces.
• Follow agency security procedures, such as not including passwords, keys, or security credentials in prompts or datasets input into GenAI technologies.
• Follow your agency’s policy on using open-source code and comply with agency licensing and attribution requirements.

To learn more about these uses and any others in which you are interested, consult with your agency’s Chief AI Officer or other agency AI governance officials.

Continuous learning and training

This is an evolving field, and as we adapt to the tools available to us, it’s important to stay up to date with relevant trainings. The Office of Management and Budget (OMB), in collaboration with the General Services Administration (GSA), has piloted free AI training for federal employees and will continue to create, adapt, and deliver federal AI training modules in partnership with OPM and informed by agency needs. As a federal employee, you can access the 2023 training here (MAX Page).

Additionally, Executive Order 14110 tasks agencies to implement — or increase the availability and use of — AI training and familiarization programs for employees, managers, and leadership in technology, as well as in relevant policy, managerial, procurement, regulatory, ethical, governance, and legal fields.

Contact your agency’s chief human capital officer or chief learning officer to learn more or join the AI Community of Practice.

AI communities of practice

If you are interested in AI or intend to use, manage, develop, or procure GenAI technology, you are encouraged to join the AI Community of Practice (AI CoP), open to any federal government employee. This community of practice regularly shares resources including centralized information around training.

Endorsement disclaimer

Links to and mention of brands, services, products, or websites external to the U.S. Government on this webpage are provided as examples and for informational purposes only; they do not constitute an official endorsement or approval by the U.S. Government, or U.S. Office of Personnel Management, of any brand, service, product, or site of the corporation, organization, or individual.