Sola Security Taps Prof. Randall Balestriero to Rethink How AI Understands Cybersecurity
Photo Courtesy: Randall Balestriero

Sola Security Taps Prof. Randall Balestriero to Rethink How AI Understands Cybersecurity

By: Ethan Rogers

Finding a cybersecurity problem is not necessarily the hardest part of resolving it. Understanding what the problem affects, how different systems are connected, and what could happen after taking action can be considerably more complicated.

That challenge is becoming increasingly relevant as enterprises look to deploy AI agents across their security operations. An agent may be able to identify a compromised identity or a misconfigured resource, but acting on that information requires a broader understanding of the environment.

Sola Security is building its research direction around this distinction. The frontier cybersecurity lab has announced that Prof. Randall Balestriero, Assistant Professor of Computer Science at Brown University and a visiting researcher at Meta AI Research (FAIR), has joined its scientific advisory board.

A collaborator of renowned AI researcher Yann LeCun, with whom he has co-authored research, Balestriero brings expertise in self-supervised learning and representation learning. His work explores how AI systems develop structured representations of the world around them.

His role at Sola will involve advising on research directions that connect foundational AI research with the operational realities of enterprise cybersecurity.

The Missing Piece in Security Automation

Security teams operate across environments where identities, permissions, workloads, and infrastructure components are closely connected. A problem affecting one component can have implications for several others.

Consider the example of a compromised identity. Identifying the account as suspicious does not answer every question a security team needs to address. The team must understand which workloads the identity can access, what permissions it holds, and how those permissions interact with other systems.

Even a straightforward remediation action, such as revoking access, requires additional consideration. Removing permissions could affect workloads or other interconnected systems.

For Sola, this illustrates a limitation in security approaches that focus primarily on individual findings. Detecting a threat is valuable, but understanding its reach and determining the appropriate next step requires a representation of the broader environment.

The company believes AI agents will need to develop this contextual understanding to make more informed decisions.

From Detecting Threats to Understanding Environments

Sola’s research direction involves exploring new types of models that can understand environments and relationships. The company sees this capability as a potential foundation for the next generation of cybersecurity agents.

Instead of relying on disconnected security findings, these systems would need a continuously updated representation of the enterprise environment. This would include relationships between identities, permissions, workloads, and dependencies.

Such a representation could give AI agents a broader basis for evaluating security decisions. They could potentially understand how one component affects another, determine what access an identity actually provides, and assess how a proposed intervention might affect the organization.

This is particularly relevant when agents move from investigation into remediation.

An agent that recognizes a compromised identity can help identify a security issue. An agent that also understands the identity’s potential reach, evaluates the operational consequences of revoking its permissions, and determines an appropriate response would represent a different level of capability.

Sola sees this progression as part of a broader shift toward security systems that can reason about the environments they protect.

Why Representation Learning Matters

Balestriero’s appointment brings a specific area of AI research into that conversation.

His work in self-supervised learning and representation learning examines how AI systems develop structured representations of the world. These concepts are relevant to Sola’s interest in models that can understand relationships and dependencies rather than simply process isolated observations.

The company wants to explore how foundational AI research can inform security systems that understand how enterprise environments function.

Guy Flechter, Co-Founder and CEO of Sola Security, sees the distinction as an important one for the industry’s future.

“Cybersecurity has spent years building systems that can identify problems, but understanding the environment behind those problems is an entirely different challenge,” said Flechter. “We believe the next generation of security agents will need something closer to a world model: an understanding of how systems connect, how changes propagate, and what the consequences of an action will be before it is taken.”

Flechter added that Balestriero’s expertise could help connect these research concepts with the company’s cybersecurity ambitions.

“Randall brings a depth of expertise in how AI systems learn representations of the world. Bringing that perspective into cybersecurity is an important step for us. We want to build agents that don’t just react to what they observe, but understand the environments they protect and can reason about the impact of their decisions.”

A Different Research Agenda for Cybersecurity

The appointment also highlights Sola’s ambition to approach AI-native security through foundational research rather than simply applying existing models to familiar cybersecurity workflows.

The intersection of AI research and cybersecurity raises questions about how systems can understand relationships, anticipate outcomes, and make decisions with awareness of their operational surroundings.

As enterprises increasingly deploy AI agents to investigate threats, manage access, and take remediation actions, these questions could become more important.

For Sola, world models offer a potential path toward addressing them. The company’s broader objective is to close the gap between identifying a security problem and understanding what should happen next.

Bringing Prof. Balestriero onto its scientific advisory board gives Sola an opportunity to explore that direction with expertise rooted in the study of how AI systems learn to represent the world.

This article features branded content from a third party. Opinions in this article do not reflect the opinions and beliefs of New York Weekly.