Enterprise AI security has entered a phase unlike any previous technology cycle, as frontier models surface software flaws that went unnoticed for decades. The result is a threat landscape expanding in two directions at once: Defenders gain new tools, while attackers gain the same leverage.
That dual dynamic is pushing chief information security officers to rethink architecture, not just tooling. Service providers working across hundreds of enterprise environments are seeing customers hesitate, waiting for the market to settle before committing to a strategy, according to Vishal Salvi (pictured), senior vice president and global head of the cybersecurity service line at Cognizant Technology Solutions Corp.
“What’s different in the last six months is the way in which AI is impacting and making a big change to the ways we look at cyber, both in terms of how AI is deployed within the cybersecurity, but also adversarial AI,” Salvi said. “If you look at it in terms of what frontier AI is doing right now, it’s able to find out vulnerabilities which humans have missed for almost three decades. All these dormant vulnerabilities are becoming visible at scale. That is creating an unprecedented risk which we have never seen before.”
Salvi spoke with theCUBE’s Dave Vellante and Rebecca Knight at Fal.Con, during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. (* Disclosure below.)
Consolidation and guardrails reshape ai security strategy
Cognizant, which recently launched secure AI services aimed at helping enterprises scale agentic systems, is pressing customers to simplify rather than accumulate tools. Fewer, deeply integrated partners shrink the blast radius when something goes wrong, Salvi noted.
“If I work with multiple tools, actually I have a bigger risk than if I work with a more simplified architecture,” Salvi said. “In cyber world, we don’t only worry about just managing and protecting, because we cannot assume that you’re not going to get breached. The key thing is how soon you can act and how soon you can contain and recover.”
Agent sprawl widens that risk further. Conservative estimates point to 50 non-human identities for every human in an organization, and possibly 100 in the future, Salvi noted. Governing them is a policy problem the industry already knows how to solve, he explained.
“You need to teach the agents the same aspect, because if you don’t, then they will go berserk and they will execute their task without these guardrails,” Salvi said. “It’s not rocket science. We know how to build policies, and we need to apply the same policies for the non-human identities.”
Here’s the complete video interview, part of SiliconANGLE’s and theCUBE’s coverage of the Fal.Con event:
(* Disclosure: TheCUBE is a paid media partner for the Fal.Con event. Neither CrowdStrike, the sponsor of theCUBE’s event coverage, nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)
