How we use AI responsibly to support your care and benefits

Before VA employees use any AI tool for your care or benefits, we conduct a detailed, multi-step review to make sure the tool is secure, respects your privacy, treats everyone fairly, and works as intended. We then continue to monitor the tools for as long as it’s in use at VA.

 

Here are the steps in our process:

 

1. Review before deployment. Before we use any AI tool, the requestor must submit the use case to the VA Chief AI Officer (CAIO) team for an independent, 2-party review before deployment. We conduct this independent review to evaluate potential risks, compliance with federal requirements, and whether the tool is appropriate for its intended purpose.
2. Ensure extra scrutiny for AI tools that impact healthcare and benefits. If an AI system could significantly affect your healthcare, benefits, or other important rights, we apply deeper levels of review. These reviews evaluate the data, potential risks, human oversight, and testing before and after deployment. The regulation that covers high-impact cases is called “Accelerating Federal Use of AI through Innovation, Governance, and Public Trust (OMB-M-25-1).
   Read OMB-M-25-1 on the White House website (PDF)
3. Complete security and privacy clearance. Like all tools we use at VA, before AI tools can access VA networks or Veteran information, they must complete an Authority to Operate (ATO). The ATO process is a rigorous security and data-protection review.
4. Conduct ongoing monitoring. VA continues monitoring AI after it is deployed to ensure it performs safely and as expected. If a tool no longer meets VA’s standards, it can be limited, suspended, or removed from use.
5. Ensure that people stay in charge. Your doctor, nurse, claims processor, or other VA employee remains responsible for decisions affecting your care or benefits. AI is designed to support their work, not replace their judgment or accountability.

We may use AI to support the people who serve you in ways like these:

 

• Helping spot something for a provider to look at, like patterns in a scan
• Speeding up routine steps, like checking that a form is complete before a person reviews it

 

AI doesn’t make final decisions, and it does not replace your healthcare provider or claims processor.

 

You remain at the center of every decision about your care and benefits; whether or not AI is involved with making that decision, we care about making the right decision in a timely manner so that you receive the care and benefits that you’ve earned. If you ever disagree with a VA decision regarding your benefits, you have the right to petition for a second opinion.

No.

 

We use AI to help your care team and claims processors do their jobs so they can serve you better. For example, AI can check that a claim is complete and filled out correctly before a claims processer reviews it. That means your claims processor can spend more time on the part that matters most—making sure your claim gets a fair and accurate decision. It also means your claim may move faster.

By law, we have a monitoring program in place for all AI tools defined as “high-impact” cases. High-Impact AI cases are those where AI outputs serve as a principal basis for decisions or actions with a legal, material, binding, or significant effect on any of these areas:

 

• Your civil rights, civil liberties, or privacy
• Your access to education, housing, insurance, credit, employment, or other programs
• Your access to critical government resources or services
• Human health and safety
• Critical infrastructure or public safety
• Strategic assets or resources, including high-value property and information marked sensitive or classified by the federal government

 

For these tools, we require an AI impact and risk mitigation plan that include testing, validation, and monitoring before being put into use. This plan defines what happens if the tool drops below an acceptable performance level. Depending on the case, we may take actions such as suspending use of the tool or removing it from service entirely until we can resolve the issue. We design these risk management practices to minimize the likelihood and impact of AI mistakes.

 

The regulation that covers high-impact cases is called “Accelerating Federal Use of AI through Innovation, Governance, and Public Trust (OMB-M-25-1).

 

Read OMB-M-25-1 on the White House website (PDF)

Our use of AI doesn’t change your rights as a Veteran. And the person who makes or approves any decision is still accountable for that decision.

 

No matter whether we used AI in your care or benefits process, you still have these rights just as you do for any decisions made without AI.

 

For health care decisions:

 

• You have the same right to appeal a treatment decision you disagree with.
  Learn more about Clinical Appeals of medical treatment decisions
• If you ever have questions about how we may have used AI in your care, you can always ask your care team. You can also connect with a patient advocate to share your concerns.
  Learn more about getting help with concerns at a VA health facility

 

For benefit decisions:

 

• You have the same right to request a decision review of any benefit decisions you believe are incorrect or if you have new evidence to share.
  Learn more about choosing a decision review option
• You have the right to work with an accredited attorney, claims agent, or Veterans Service Organization (VSO) representative. These trained professionals can help you file a claim or request a decision review.
  Learn how to get help from a VA accredited representative or VSOA

Protecting your personal and health information is one of our top priorities. AI tools must meet the same strict privacy and security requirements as every other tool we use to serve Veterans and their families.

 

We use several layers of protection to keep your information safe:

 

• Zero trust architecture. Since 2022, we’ve followed a zero-trust security model. We don’t automatically trust any user, device, or system. Instead, we verify every request to access data first.
• Security review before deployment. Any tool that accesses the VA network must complete our Authority to Operate (ATO) process. The ATO process is a rigorous security and data-protection review.
• Minimal access, by design. We give each AI tool access to only the information the tool needs to perform its approved task, nothing more.
• Federal compliance. We follow all federal data privacy and security regulations that apply to Veterans’ information, whether a person or a tool uses that information.

 

Learn more about our VA cybersecurity strategy

 

Read our VA privacy policy

To help protect Veterans, VA employees have access to training on how to use AI safely, responsibly, and appropriately. They remain personally responsible for every decision they make, whether they use AI to help them or not.

 

Here are some examples of guidance and training we provide to VA employees:

 

• Personal accountability. VA employees remain responsible for the decisions they make or approve, even when AI helps inform their work.
• Responsible use. We provide employee training on when they can use AI, when they can’t, and how to handle sensitive information safely.
• AI governance. We provide training on how we review, approve, monitor, and retire AI tools throughout their lifecycle.
• AI fundamentals. We provide training to help all VA employees build baseline understanding of what AI can and cannot do.
• Tool-specific training. We provide in-depth training on the approved AI tools available at VA and how to use them appropriately.
• Generative AI guidance. We provide employee guidance on the safe and appropriate use of tools such as AI chatbots and writing assistants.

 

This training and guidance helps ensure that AI supports, not replaces, the judgment and expertise of the people who care for and serve you.

We’re always searching for ways to better support our mission to provide high-quality, trustworthy care and services to Veterans. Here are some examples of how we’re using AI today to improve the speed, consistency, and quality of support to Veterans:

 

• AI is helping us VA process significantly more benefits requests far faster by conducting the compliance checks we used to have humans perform. Our benefits experts can now focus more attention on the actual content of your requests so you can get what you’ve earned, sooner.
• FDA-approved AI image analysis tools are helping our physicians detect more potential cancers earlier. This earlier detection improves Veteran outcomes and reduces treatment severity.

 

Want to learn about more ways we’re using AI to help you?

 

Review our AI inventory