Security in the Age of AI with Anand Oswal

At this year’s RSAC Conference, one theme loomed large: AI isn’t just a tool anymore—it’s a battleground. Industry veteran Anand Oswal discussed how AI is reshaping both sides of the cybersecurity equation: It’s amplifying the speed and scale of attacks while simultaneously offering new ways to fight back.

The complexity of securing AI applications is quickly outpacing the capabilities of traditional, point-based security tools. Organizations today are expected to secure everything from LLM-driven agents and personalized AI assistants to infrastructure, models, plugins and datasets—all while ensuring safe and compliant use of browser-based AI tools by employees.

And that’s not even scratching the surface. Oswal pointed to emerging risks that sound more like science fiction: poisoning agent memory, prompt injection attacks, and model-level denial of service. The implication? AI security isn’t about scanning a model and calling it a day. It requires rethinking posture management, red teaming, runtime protection and access control—across both human and autonomous actors.

There’s also the reality that most organizations are drowning in tools. Many are juggling a dozen or more solutions that don’t integrate well, leading to fragmented policies and operational fatigue. What security leaders are asking for now isn’t more alerts—they want outcomes. Fewer interfaces. Unified workflows. Simpler decision-making.

Perhaps the most optimistic takeaway: AI might actually make that possible. If done right, it can unify visibility, enforce consistent controls across environments, and preempt zero-day attacks. As Oswal put it, AI may finally make security “solvable”—not perfect, but achievable, at least for those willing to abandon fragmented approaches in favor of more integrated, adaptive architectures.