Radiant Logic helps MSSPs track AI agent identity risk

The World Economic Forum (WEC) laid out the complicated nature of AI when it comes to cybersecurity, serving as a powerful defense against adversaries as well as an attack vector for hackers and a tool that those bad actors can use in their campaigns.

The rise of agentic AI only adds complexity to the equation, representing what the WEC called a “breakthrough and a burden.”

“It's cyber defense at machine speed,” the organization wrote. “On the other hand, these same capabilities can be weaponized. ... When both attackers and defenders operate at microsecond intervals, the nature of cyber conflict transforms. The line between shield and sword has never been thinner.”

Sébastien Faivre is chief product officer for Radiant Logic, one of many vendors looking to help enterprises manage the intersection of AI agents and identity. A problem, according to Faivre, is that agents pull apart the assumptions that identity security has relied on for decades.

“A human identity is stable: one person, one set of credentials, predictable behavior,” he told MSSP Alert. “Agents are the opposite. They’re created and destroyed in seconds, act at machine speed without a human in the loop, and often inherit broad permissions no one is governing. Many get spun up outside any central process, so security teams never even see them.”

Complex Security Risks

Faivre added that “the risk is the combination of scale, autonomy, and entitlement. The adversary doesn’t break in; they log in, and agents hand them a new, poorly defended set of credentials to exploit.”

Radiant Logic, last week, unveiled new capabilities for its RadiantOne Identity Data Platform that will allow enterprises and MSSPs to inventory agents, control them, and generate continuous risk scores for them, regardless of who built them, where they were built, or which platform they run on. According to Radiant Logic, the platform can monitor the behavior of each agent in real time, recalculate the agents’ risk as posture and privilege, and quickly get remediation underway when needed.

The vendor plans to demonstrate the features at the Identiverse 2026 show starting June 15 in Las Vegas.

Fast Adoption of Agentic AI

Radiant Logic’s new capabilities come as corporate adoption of AI agents is increasing rapidly. Vasu Jakkal, corporate vice president for Microsoft Security, wrote earlier this year that 80% of the Fortune 500 companies use AI agents.

That said, most organizations aren’t ready for the security risks they bring, Faivre said.

“Adoption is racing ahead of governance,” he said. “Teams are deploying agents to capture productivity gains while security and risk functions are still trying to answer basic questions: how many agents do we have, who owns them, what can they access, and what happens when one is compromised? The key risks we see are unmanaged proliferation, excessive privilege, unclear ownership, and no consistent way to judge which agents actually pose the greatest risk.”

The Burden on MSSPs

For MSSPs and MSPs, they bear the burden twice: once for themselves and then again across every client they are hired to protect, Faivre said.

“Each client adopts agents on its own timeline with its own gaps, creating a huge, fragmented attack surface,” he said. “Risk scoring changes the economics. Instead of treating every agent identity as equal, providers can focus on the ones that represent real exposure, report posture to clients in concrete terms, and scale their service without scaling headcount.”

For an MSSP that is managing many tenants, “being able to consistently answer ‘which agents should we worry about, and why’ is the difference between offering agent security-as-a-service and being overwhelmed by it,” Faivre said.

The new agent visibility capabilities are the latest effort by Radiant Logic to push more AI features onto its RadiantOne platform. In December 2025, the company announced AI-driven remediation and access capabilities to address the thorny problem that while identity risks can be quickly detected, fixing them can often take too long.