Artificial intelligence is opening new possibilities for enterprise cybersecurity, particularly as organizations explore agents capable of investigating threats, managing access, and taking remediation actions. Yet increasing the capabilities of these systems introduces a fundamental question: how can an AI agent make an informed decision if it does not fully understand the environment in which it operates?
Sola Security is positioning this question at the center of its research agenda. The frontier cybersecurity lab has appointed Prof. Randall Balestriero, Assistant Professor of Computer Science at Brown University and a visiting researcher at Meta AI Research (FAIR), to its scientific advisory board.
Balestriero has collaborated with renowned AI researcher Yann LeCun and co-authored research with him. His work in self-supervised learning and representation learning explores how AI systems develop structured representations of the world around them.
Through the appointment, he will advise Sola on research directions that connect foundational AI research with the operational requirements of enterprise security.
The move reflects a broader ambition to develop security agents that can understand relationships between systems, evaluate potential outcomes, and make decisions with greater awareness of their surroundings.
Why Security Agents Need More Than DetectionThreat detection is an essential part of cybersecurity, but identifying a suspicious identity or misconfigured resource does not necessarily provide enough information to resolve the underlying problem.
Enterprise security teams must understand how individual components relate to one another. An identity, for example, may have permissions that provide access to multiple workloads. If that identity becomes compromised, determining the extent of the exposure requires an understanding of those relationships.
The remediation process introduces another challenge. Revoking the identity's permissions might eliminate an immediate risk, but security teams also need to understand what could break as a result.
These are not isolated technical questions. They require an understanding of dependencies and the potential consequences of changes across an interconnected environment.
Sola believes that AI agents need a continuously updated representation of enterprise environments to address these challenges. This would allow them to move beyond processing individual security findings and develop a more comprehensive understanding of the systems they protect.
A Different Approach to AI Decision-MakingThe company's research direction centers on the potential of new types of models that understand environments and relationships.
Applied to cybersecurity, these principles could help AI agents build representations of how identities, permissions, workloads, and infrastructure dependencies interact. With this context, agents could potentially evaluate security interventions based on their broader operational implications.
For example, an agent investigating a compromised identity would need to understand more than the identity's status. It would need to determine what resources the identity can access, how those permissions connect to other systems, and what consequences might follow from removing them.
This approach could help establish a more informed decision-making process for AI-driven security operations.
Rather than simply responding to an alert, agents could evaluate potential actions, identify downstream effects, and determine appropriate responses with a deeper understanding of the environment.
Sola sees this capability as an important foundation for the next generation of cybersecurity agents.
The Research Behind Sola's DirectionBalestriero's appointment brings expertise in two areas of foundational AI research: self-supervised learning and representation learning.
His work examines how AI systems develop structured representations of the world around them. These research areas connect with Sola's interest in building models that can understand the relationships and dependencies present in enterprise environments.
As a visiting researcher at Meta AI Research (FAIR) and an Assistant Professor of Computer Science at Brown University, Balestriero brings an academic perspective to the company's efforts. His previous work alongside Yann LeCun further connects the appointment to research exploring how AI systems learn and represent information.
Guy Flechter, Co-Founder and CEO of Sola Security, described the distinction between identifying security problems and understanding the environments behind them as an important challenge for the industry.
"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.”
He also explained why Balestriero's expertise is relevant to Sola's 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."
Building Toward Context-Aware CybersecurityThe appointment is part of Sola's ambition to approach cybersecurity as a research-driven field at the intersection of foundational AI and enterprise security.
Rather than focusing exclusively on adapting existing models to established security workflows, the company is exploring how advances in AI research could support systems capable of understanding relationships, dependencies, and potential outcomes.
As AI agents become more involved in enterprise security operations, their ability to understand the environments they protect could influence how effectively they investigate threats and carry out remediation.
For Sola, the opportunity lies in developing systems that can bridge the gap between recognizing a security issue and understanding the right course of action.
Written by the TVC editorial team.