In the generative AI gold rush, it’s easy to forget that no amount of modeling magic can compensate for low-quality data. As enterprises race to deploy artificial intelligence, they often overlook the less glamorous but utterly essential work of securing, governing and preparing their data foundation. That omission is more than an operational oversight; it’s a strategic liability, especially when it comes to data protection for AI.

“The lifeblood of AI is data,” said theCUBE Research’s Scott Hebner (pictured, right). “There’s no AI without an … information architecture. The data is the critical component of any AI system, and so that data has to be of the highest quality. It has to have integrity built into it on how it’s processed, how it’s used, and most importantly, it’s got to be protected. Both regulatory and from various threats that may be out there. So any good long-term AI strategy is going to start with your data layer, and that includes data protection.”

Hebner spoke with theCUBE’s Christophe Bertrand (left) at the Data Protection & AI Summit, during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. Their discussion explored the data trust gap, the evolving risk landscape around agentic AI and the foundational role of protection in building resilient and intelligent AI infrastructure.

Why data protection for AI is the limiting factor

Most enterprise data is neither protected nor ready for AI. The majority of enterprise data is effectively sidelined due to risk concerns or a lack of governance, according to Hebner. Without strong data protection for AI, even advanced models operate on a narrow slice of usable information, limiting both quality and trust.

“If you look at the vast majority of the data that an organization sits on, the proprietary enterprise data, very little of it is actually used today in general, and even less is used in their AI. And when you kind of dive a little bit deeper into these AI projects, you find out that it’s because they’re not sure how to protect it. It’s a risk assessment. The data’s there, but it’s not ready to be used for AI. And we estimate some 95% of an enterprise’s data is just simply not ready, and protection’s a big part of that.”

Effective protection makes data not only secure, but also ready for AI use and reuse at scale. That’s why data protection for AI must be treated as a design requirement, not a post-deployment afterthought, according to Bertrand.

“AI is important to you as a workflow, it’s important to you as a workload, and important to you as a business,” Bertrand said. “For this reason, anything that is part of AI infrastructure has to be protected. I think that’s the baseline. Don’t make it a second thought. It is actually a design requirement.”

As AI agents and automation generate new data at unprecedented speeds, governance must keep pace. Data protection for AI plays a pivotal role in that process, ensuring continuity, integrity and trust as systems evolve, according to Hebner.

“I think [that AI] is going to learn from that data and start to understand what is really quality data, what is protected and what is not,” he said. “Then, it will learn from that and come back and do it better next time. As the system builds over time, it becomes better and better at what it does. I do think building in an AI … architecture that underpins governance and trust management, those frameworks including protection and regulatory compliance and corporate policy is really … sometimes you wonder if it becomes table stakes. It’s just something you have to do; otherwise, how are you going to keep up with all this?”

Here’s the complete video interview, part of SiliconANGLE’s and theCUBE’s coverage of the Data Protection & AI Summit:

Photo: SiliconANGLE