Federal AI: Trust, Governance and Mission Outcomes

August 12, 2026

In the summer of 2026, the SAS Innovate Tour arrived in Washington, DC, at Convene Hamilton Square. 50 years have gone by since SAS launched a contract from the Department of Agriculture just down the street.

It seemed appropriate to sit down with Caroline Baldwin, the Director, SAS Federal, to discuss how SAS delivers trusted AI for the federal government.

Everybody reading this has seen reports of AI giving hallucinations. It is one thing to have a bad response for a doughnut company in Chicago; quite another to have the Department of War base decisions on bad data.

Caroline Baldwin brings a refreshing splash of reality to the talk of applied AI to the federal government. Instead of evaluating the nuances of AI harnessing, she makes a blunt statement, “If government can’t apply AI to outcomes for an agency or improve American lives, it’s just a science project.” Finally, we have a practical opinion.

SAS has been working with sensitive federal data for fifty years and has the experience to ingest accurate data and use it to produce practical results for federal agencies.

We have heard of the “black box” of AI. Don’t ask for sources. In Baldwin’s experience, an AI system must be able to produce sources for decisions that will be held up in court.

During the interview, she mentioned that SAS can leverage synthetic data to produce more accurate models.

She also describes how SAS partners with federal agencies to solve complex problems, adapt commercial innovations, and accelerate adoption. As AI evolves, agencies must balance enterprise strategies with mission-level needs while continuously updating models and maintaining trust.

"If government can't apply AI for an outcome for the agency or to improve American lives, it's just a science project

Caroline Baldwin SAS

SAS has been working with sensitive federal data for fifty years and has the experience to ingest accurate data and use it to produce practical results for federal agencies.

We have heard of the “black box” of AI. Don’t ask for sources. In Baldwin’s experience, an AI system must be able to produce sources for decisions that will be held up in court.

During the interview, she mentioned that SAS can leverage synthetic data to produce more accurate models.

Social

email

Policy

Federal Tech Podcast

Download the Federal Tech Podcast Media Kit