Bob Gourley and Matt Devost, co-founders of OODA LLC, recently made an appearance on Federal News Radio’s What’s Working in Washington.

In the interview, Bob Gourley explained how the problem of breach fatigue may lead to the national security implications of certain data breaches being ignored: “[W]e’ve become numb to all these attack notifications, breach notifications. There’s just millions of records stolen all the time, and so, how do you sort it out, and realize what is important or not?” This is a crucial question, since recent incidents like the Marriott data breach and the cyberattack on Tribune Publishing have implications for national security, even though the targets were private companies.

Matt Devost elaborated on the security aspects of the proliferation of A.I. technology, such as A.I. bias: “When A.I. programs are written, or algorithms are written, they kind of inherit the bias of the authors. … [W]e’ve seen this over and over again in the market, where machine learning or A.I. is introduced, and as it continues to build upon the learning in which it was programmed, those biases start to get enhanced.” This makes data integrity essential: “We need to make sure that these A.I. and machine learning systems can’t be influenced by external actors, where we see increasingly the use of machine learning to process signals from external sources.”

Matt Devost also pointed out that if security is not built into A.I. from the start, people might lose control over the development of this technology in the future: “[A]s we implement machine learning, we might reach a point where we don’t really even understand how the technology is working, so let’s start with a very strong security foundation at the beginning. Let’s build in the best practices that we know, let’s secure these platforms. Use that as the baseline.”

Read more: OODA: both a philosophy and a company on the rise