Can Technology Be Trusted to Act on the Data It Collects?

October 15, 2026 - 01:42

The rapid expansion of artificial intelligence into health care has raised a question that goes beyond data gathering. If technology can help collect information, can it also be trusted to act on it?

Dr. Haider Warraich points to a clear gap between promise and oversight. "There's no real generative AI technology or agentic AI technology that has ever received FDA authorization," he said.

That statement matters because agentic systems are designed to do more than summarize or suggest. They can make decisions, trigger actions, and operate with limited human input. In a clinical setting, that shift changes the stakes. Collecting a patient's history or flagging a lab result is one thing. Choosing a treatment path or adjusting a medication dose is another.

Regulators have approved many tools that use machine learning for narrow tasks, such as reading images or detecting patterns in scans. Those tools typically work inside strict boundaries and undergo review before reaching patients. Generative and agentic systems behave differently. They adapt, produce open-ended outputs, and can chain actions together in ways that are harder to predict.

Warraich's comment highlights a regulatory vacuum. Hospitals and startups are already testing these systems, sometimes in pilot programs that fall outside traditional device review. Without authorization, there is little standardized evidence about safety, bias, or failure modes.

The question is not whether the technology can act. It is whether anyone has proven it should. Until that evidence exists, trusting AI to act on the information it collects remains an open risk rather than a settled fact.