At the RSA Conference in San Francisco, Relyance AI CEO and co-founder Abhi Sharma outlined a fundamental shift underway in enterprise technology: the transition from deterministic software systems to non-deterministic, AI-driven environments.
In this new paradigm, traditional assumptions about security, governance, and trust no longer hold. Sharma argues that organizations must rebuild their approach from the ground up, centering on data as the core risk vector. Relyance AI offers an AI-native data security platform, designed not as an add-on layer but as an embedded, continuous system that tracks data across its full lifecycle.
With the introduction of LYO, an autonomous data defense engineer, the company aims to move security from reactive oversight to proactive, AI-driven orchestration—bringing trust, compliance, and governance into direct alignment with the pace and complexity of modern AI systems.
Core Takeaways
- Understanding AI-native: AI-native organizations are built with AI as a foundational operating model, not an add-on, enabling dramatically faster development cycles and new forms of system design driven by agent orchestration rather than manual coding.
- Deterministic to non-deterministic: The move from deterministic to non-deterministic software fundamentally breaks traditional security assumptions, requiring a complete rethinking of how data is governed and monitored.
- Focusing on data risk: Relyance AI differentiates itself by focusing on data risk as the core of AI security, tracking the full lifecycle of data through AI Data Journeys and mapping exposures through a dynamic data exposure graph.
- LYO agent: The company’s LYO agent represents a shift from tools to autonomous systems, acting as a data defense engineer that not only identifies risks but recommends and executes remediation in real time.
Key Quotes
AI-Native vs. AI-Bolted Systems
“I think the central idea is, you define your operating model with AI being a key and essential component of your operations—not just in one department, but across the board. It’s not something that’s added on later; it’s built that way from the start. That changes everything about how systems are designed and how work gets done.
“The clearest example is in engineering. In traditional environments, developers still write code, maybe with AI assisting them. But in AI-native companies, engineers are increasingly orchestrating agents that do the work for them. They’re directing systems rather than building everything line by line, and that represents a fundamental shift in how software is created.”
Why Traditional Security Models Break
“We used to live in a world where software was deterministic—you wrote code, tested it, deployed it, and it behaved exactly as expected. But now we’re dealing with systems that are inherently non-deterministic. A single change can create thousands of unpredictable data flows and outcomes.
“That breaks every assumption we had about security. Those assumptions were built around systems that changed at human speed and behaved predictably. Now we have systems changing at machine speed, producing outcomes we can’t fully anticipate. That’s why we believe security has to be rebuilt from the ground up.”
Data as the Core of AI Risk
“Our fundamental belief is that AI security risk is really data risk. If you can trace the full journey of data—how it’s created, transformed, copied, and used—you can actually govern and secure AI systems effectively.
“The problem with traditional approaches is that they react after something happens. What we’re doing instead is staying in sync with the flow of data as it moves through systems. That allows organizations to understand risk in real time, rather than trying to reconstruct what went wrong after the fact.”
From Tools to Autonomous Security Engineers
“LYO is not just a better scanning tool or a more advanced firewall. It’s an AI-native data defense engineer that augments your security team. It identifies issues, explains them in context, and increasingly takes action to resolve them.
“What we’re really doing is shifting from a world of tools to a world of autonomous systems that work alongside humans. The goal is not to replace people, but to eliminate the 80% of repetitive, clerical work so they can focus on the 20% that truly drives impact. That’s where the real transformation happens.”
The Future of Trust in an AI-Driven World
“We’re entering a new modality of trust where AI reshapes not just tools, but processes and the roles people play. Today, security and compliance functions tend to lag behind innovation, trying to catch up after systems are already deployed.
“What we’re aiming for is to bring trust, security, and compliance directly into the moment of creation. If engineers can understand these constraints while they’re building, rather than after deployment, you fundamentally change the equation—from reactive control to proactive alignment with innovation.”
