As AI spreads and corporations' security risks grow, security company Palo Alto Networks advised that security should be embedded from the AI service planning stage.
Palo Alto Networks said on the 22nd that it participated as a global sponsor at AWS Summit Seoul 2026 and shared an integrated security strategy and a "Secure AI by Design" approach for the AI era.
Kim Beom-soo, a professional at Palo Alto Networks, took the stage as a speaker on the 21st and gave a presentation on "Secure AI by Design." Kim explained the main AI security risks corporations face as external AI tools, AI agents, and custom AI applications spread—including model vulnerabilities, exposure of sensitive data, and agent control issues—and presented an integrated, platform-based security strategy to address them.
Palo Alto Networks emphasized that AI security should not remain a simple after-the-fact response and must be embedded throughout the entire lifecycle of AI services—planning, development, deployment, and operations. It said corporations in particular should manage, in an integrated way, risks such as shadow AI, data leakage, degraded model integrity, and agent malfunctions that can arise as they use Generative AI, software as a service (SaaS), AI agents, and in-house AI applications.
It also said that because modern AI workloads are inherently tied to cloud environments, cloud security should be redefined based on an approach that protects AI from the design stage. If not embedded proactively, organizations could be exposed to new and complex attack surfaces that are hard to detect with traditional perimeter-based cloud security tools.
Meanwhile, Palo Alto Networks is supporting corporations' AI operations with Prisma Browser, which helps govern external AI use, and "Prisma AIRS," which supports AI application development, model integrity, and agent security.
Prisma Browser provides visibility and control over Generative AI and SaaS usage within corporations to help reduce the risk of sensitive data leakage and shadow AI use. Prisma AIRS offers functions such as model integrity validation, prompt and data protection, and agent behavior control throughout the AI application and agent development lifecycle.
Park Sang-kyu, head of Palo Alto Networks Korea, said, "AI is rapidly accelerating corporations' productivity and innovation, but as AI is deployed across cloud environments, complex new risks are also emerging that are difficult to address with existing security frameworks," adding, "For corporations to innovate safely, they need to shift to a 'Secure AI by Design' model that embeds security guardrails from the early stages."