The Responsible Artificial Intelligence for Veterans Act (H.R.9862) — having an outside research center evaluate AI that is already running, after the fact

1Evaluating after deployment, not before

Most bills about AI attach conditions to what is about to be introduced. What this one addresses is AI already running inside the Veterans Health Administration. Systems deployed, in pilot, or in active development are to be evaluated by an outside research center. Rather than creating an approval process, it puts what is in use through an inspection afterwards.

The drafting assumes a situation in which AI entered clinical settings first and evaluation follows behind.

2What prioritizing at least five accomplishes

The bill directs that at least five systems be prioritized — those deployed at scale or presenting elevated clinical, operational or patient safety risk. Rather than evaluating everything at once, the instruction is to start where the effect is largest. Fixing a number brings the work down to a size that can actually be begun, which reads as care taken so the evaluation does not remain a plan. How systems not prioritized are handled does not appear in this portion.

3The evaluation reaches whether things connect

Matters for evaluation cover not only accuracy or effectiveness but interoperability with electronic health records, medical devices, pharmacy systems and other platforms. AI in a clinical setting does not run alone; it becomes usable only once it meshes with existing records and equipment.

That the adequacy of technological infrastructure supporting deployment and scalability is also an item follows the same logic: performing well means nothing if it will not sit on the infrastructure of the floor. Among the 114 US AI-related bills this site holds as of 2026-08-31, this is the one referred to the Committee on Veterans Affairs.

4Deployment first, evaluation after

Most AI bills attach conditions to what is about to be introduced. What this bill addresses is AI already running in the Veterans Health Administration.

It reads as written for the situation in which AI enters clinical practice first and evaluation follows. The evaluation is to prioritise at least five systems that are widely deployed or carry high clinical, operational or patient-safety risk. What the evaluation covers includes not only accuracy and effectiveness but interoperability with other platforms such as electronic health records, medical devices and pharmacy systems.