Securing the Next Frontier: Why Agentic AI Needs Standards Now

Artificial intelligence has entered into a new niche—one defined not just by large language models, but by autonomous agents that can reason, act, and collaborate across digital ecosystems. These agentic AI systems promise transformative efficiency, from automated customer support to autonomous research assistants. Yet with new power comes new risk. The same autonomy that enables AI agents to take action also opens the door to misuse, manipulation, and adversarial control.

Industry leaders are sounding the alarm: without clear standards, AI agents may replicate the early days of the internet, when protocols like SSL and HTTPS were only introduced after breaches and fraud had already eroded trust. As Rakshith Aralimatti, a standard-setter for agentic AI security and co-lead of the OWASP Agentic AI Security initiative, explains, “We cannot afford to repeat history. Standards for AI agent security need to be built before attacks scale, not after.”

An Untapped Market in AI Security

The market has largely focused on building smarter, more capable AI models, often overlooking the security layer that governs their actions. This oversight has left agentic AI security an underdeveloped but potentially massive ecosystem. Analysts estimate that by 2030, enterprises could spend tens of billions annually on agent governance and security tools—much like the cybersecurity market that grew around securing internet protocols.

Rakshith has been at the forefront of this shift. At Palo Alto Networks, he pioneered the first-to-market AI agent security framework, embedding safeguards for agent-to-agent communication and memory protection. The result was not only enhanced resilience for enterprises but also a roadmap for how security could become a competitive differentiator in AI adoption.

“Capabilities without guardrails are liabilities,” he notes. “Enterprises will only trust AI agents when they can prove secure behavior—identity, authorization, and compliance must be baked into the architecture.”

Emerging Standards and the Role of OWASP

The open-source community is moving quickly to fill the gap. OWASP’s Agentic AI Threats and Mitigations project, co-led by Rakshith, is shaping the first public catalog of risks and defenses for AI agents. From prompt injection to rogue-agent coordination, the initiative highlights how traditional cybersecurity practices must be reimagined for AI-driven environments.

Model Context Protocol (MCP) hardening is another area gaining momentum. Originally designed to standardize how AI agents communicate with external tools, MCP now requires additional layers of authentication and misuse prevention. By spotlighting these vulnerabilities, Rakshith and his peers are helping ensure that MCP evolves as a secure foundation, not a weak link.

The momentum is beginning to show. Rakshith was a featured speaker at the RSA Conference 2025, where he argued that AI agent security deserves the same urgency as past cybersecurity revolutions. His message resonated: security frameworks for AI are not optional add-ons—they are prerequisites for enterprise adoption at scale.

Building the U.S. Standard for AI Security

Just as the U.S. once established internet security standards that enabled e-commerce to flourish, today it faces a similar inflection point with AI. Without national guidelines, enterprises risk a patchwork of inconsistent practices that attackers can exploit. With them, the U.S. could establish itself as a leader in trustworthy AI deployment, giving American enterprises a competitive advantage in the global market.

Rakshith envisions a layered approach: OWASP and open-source groups define baseline risks, enterprises adopt and extend those standards in production, and regulators enshrine best practices into compliance frameworks. “The U.S. has an opportunity to lead by setting the rules of the game,” he emphasizes. “If we define secure AI standards now, we not only protect users but accelerate innovation by creating trust in the ecosystem.”

A Defining Moment for AI Agents

The rise of agentic AI is inevitable. The question is whether its future will be marked by trust or turbulence. By driving early frameworks through OWASP, demonstrating enterprise-ready security at Palo Alto Networks, and shaping industry dialogue at RSA, Rakshith Vijayakumar Aralimatti is helping ensure the former.

The stakes are high, but so is the opportunity. AI agents could one day become as ubiquitous as web browsers, automating work across industries. But just as browsers needed secure protocols to become trusted tools, agents need robust security standards to achieve their potential.

As Rakshith puts it: “If AI agents are to become the backbone of digital work, securing them is not just a technical challenge—it’s a societal obligation.”