The warning bell is sounding for cybersecurity, as machine identities flood the system faster than companies can secure their data.

At the annual Nvidia GTC event, Nvidia Corp. CEO Jensen Huang introduced a series of open-source tools based on OpenClaw, an open-source AI agent. The main tool, NemoClaw, is designed to safeguard OpenClaw’s outputs, but security experts remain concerned.

“Everything that Jensen announced last week at GTC increases the surface area and the attack surface and the risk,” said Dave Vellante (pictured, right), chief analyst for theCUBE Research. “It’s all we talked about at [RSAC 2026]. Machine identities now outnumber human identities like 80,000 to one. Google … shared with me that they have discovered more than 800 of the OpenClaw skills that are downloadable are straight out malware.”

On the latest episode of theCUBE Pod, Vellante and John Furrier (left), executive analyst for theCUBE Research, discussed why the cybersecurity sphere needs to undergo a drastic shift. They also explored the proliferation of machine identities, as well as the relationship between cloud-native and AI-native technology.

RSAC was the canary in the coalmine for agentic AI

The mood at RSAC, the annual cybersecurity conference, was ominous, according to Vellante, who called it a “canary in the coal mine.” Agentic AI, if not addressed in a timely manner, could be disastrous for cybersecurity, leading to an influx of machine identities that are increasingly difficult to track and secure.

“Shadow AI was probably the number one or number two topic at RSAC this year,” Vellante said. “CISOs are really, really concerned about it. No question that the fear factor is very high. The other topic was there are a lot of unknown unknowns. What’s the first line of defense now? Is it identity? Is it agents? It’s shifting.”

Companies will soon need layered security throughout their infrastructure in order to combat agentic AI-powered attacks. Enterprise is already concerned about getting a return on investment for AI, Vellante and Furrier pointed out, and the threat posed by agentic AI is not reassuring.

“Everyone last year was kind of drinking the Kool-Aid,” Furrier said. “Three years ago it was: Agents are bullshit. Last year, ‘Oh, agents are good. We can do end-to-end workflows.’ This year, it’s like, ‘Oh shit, agents are bad.’ Full circle. It’s [enterprise] going from humans to autonomous systems and from reactive to real-time machine speed.”

Machine identities are the new Wild West

Machine identities are untreaded territory for cybersecurity practitioners. Instead of humans talking to humans, or agents talking to humans, security now entails agents talking to agents. And if agentic AI is the cause of the problem, then it might also have to be part of the solution, according to Furrier.

“Humans think nondeterministically,” he explained. “But the hackers are using the agents and they’re non-human and they can be non-deterministic. There’s a whole architectural shift. This is a game-changer for the stack because that’s decades, generations of security principles completely being upended.”

Everyone knows that agentic AI is a risk to cybersecurity but opinions are divided on what to do about it. Most companies have yet to fully adopt AI into their stack and implementing machine identities faster than they can be controlled would be a recipe for disaster, Vellante believes.

“Start with governance,” he said. “If you don’t have a strong governance structure in place and a strong culture of security and governance and compliance, what’s going to happen? You’re going to get shadow AI just like we have shadow IT, like we had shadow big data, except this time it’s way more dangerous.”

Cloud-native meets AI-native

KubeCon 2026 in Amsterdam gave insight into how cloud-native practitioners are responding to the demands of AI inference. Cloud-native infrastructure has become a substrate for AI-native technology, providing the necessary governance to keep models and agents in check.

“You’re starting to see the formation of this new architecture where the cloud-native mojo is coming to the table, but not in the leadership way of driving AI-native,” Furrier said. “It’s the classic standing on the shoulders of giants mentality, which is beautiful because open source is booming.”

The AI sector still needs an interface or framework to control the growing number of machine identities, according to Furrier, and no clear leader has emerged yet. As a “system of intelligence” composed of machine identities develops, enterprises will need to figure out where it lies in relation to their data, Vellante added.

“If you’re going to have a digital representation of your enterprise and having agents take action, you’ve got to have transactions,” he said. “The locality of this system of intelligence or the cognitive surface is really important. Enterprises are going to be operating … that semantic layer, that transaction layer, that’s all going to be done close to the data.”

Watch the full podcast below to find out why these industry pros were mentioned:

Jim Kavanaugh, CEO of World Wide Technology
Michael Dell, chairman and CEO of Dell Technologies
Peter Tuchman, stock trader, Einstein of Wall Street
Bob Pisani, former CNBC correspondent
Jensen Huang, president, co-founder and CEO of NVidia
Lena Smart, Ambassador – AIUC-1
Jon Oltsik, principal analyst for cybersecurity at SiliconANGLE Media
Adi Shamir, Israeli cryptographer and inventor
Bipul Sinha, co-founder, chair and chief executive officer of Rubrik
Pat Gelsinger, former CEO of Intel
David Floyer, analyst emeritus at theCUBE Research
George Gilbert, principal analyst at theCUBE Research
Danny Brickman, co-founder & CEO of Oasis Security
Ali Ghodsi, co-founder and chief executive officer of Databricks
Ben Horowitz, general partner at Andreessen Horowitz
Pete Hegseth, United States Secretary of War
Nir Zuk, founder and CTO of Palo Alto Networks

Here’s the full episode of this week’s theCUBE Pod:

Don’t miss out on the latest episodes of “theCUBE Pod.” Join us by subscribing to our RSS feed. You can also listen to us on Apple Podcasts or on Spotify. And for those who prefer to watch, check out our YouTube playlist. Tune in now, and be part of the ongoing conversation.

Photo: SiliconANGLE