Securing Agentic AI and Citizen Development Starts with Policy and Standards

Introduction

At Zenity, we're deeply committed to helping our customers harness and embrace the power of Agentic AI and citizen development platforms by getting these powerful tools into the hands of business users everywhere, responsibly. Our mission goes beyond product; we’re working to foster a secure, AI-driven world by supporting the development of compliance frameworks such as updates to the NIST CSF 2.0 and AI RMF, contributing to public policy via our partnership with OpenPolicy, sharing research done by our Labs team, and empowering the broader security community with core contributions to efforts like MITRE ATLAS, the OWASP Top 10 project for low-code/no-code development, and the OWASP GenAI Agentic Security Initiative.

However, Agentic AI largely lacks defined centralized legislative policies and security standards (legislative or otherwise) - which the market is anxiously seeking from subject matter experts that are actively engaged, using, and responsible for the security of these types of systems. Given Zenity’s expertise in securing agentic AI for the world’s leading enterprises, we have a unique opportunity, and as we see it, responsibility, to bridge this policy gap. The absence of well-defined security standards, coupled with increasing demand from enterprises and regulators, makes now the time to act.

As the U.S. government refines its AI strategy and the EU accelerates the development of AI compliance frameworks, policymakers are actively seeking industry expertise. As an active member of the security community, Zenity’s experience allows us to contribute in the efforts to shape these discussions, ensuring security remains at the forefront of AI adoption.

The Three Strategic Pillars We’re Evangelizing

To guide our policy and education efforts, we’ve distilled our mission into three strategic themes:

1) Citizen Development of AI Agents is Transformational

Much of today’s Agentic AI is being built by the average business user, sometimes known as citizen developers. Technologies such as low-code/no-code (LCNC) and ‘Vibe Coding’ platforms have democratized development, empowering citizen developers to create powerful applications, automations, and now AI agents without a technical background. This new paradigm has unlocked massive operational gains across industries, enabling users to be more efficient, effective, and productive by bringing their line of business knowledge to the development process. However, in doing so, the citizen development revolution brings unique security challenges including Shadow AI, Shadow App Development, and a wide swath of vulnerabilities that can go undetected with traditional security controls and knowledge. In order to support the continued adoption of agentic AI, we must also establish strong security foundations for citizen development and LCNC platforms. Innovation and security must go hand-in-hand.

2) Defense in Depth is Essential

Securing Agentic AI requires more than just vulnerable prompt disclosure, it necessitates defense in depth. While some risks, like prompt injection, may not be eliminated entirely, security teams can mitigate its risks through layered security approaches. Organizations must move beyond solely blocking at the perimeter, and instead focus on the behavior of agents in order to monitor for abuse and abnormalities and defend at every layer of the organization.

3) Agentic Insider Threats Are a New Paradigm

Traditional security models have long focused on applications, machines, and human users that exist across the cloud and on-premises resources. Agentic AI introduces a new dimension; as they mix elements of all the above, yet are nuanced in new ones. The autonomous nature of agents makes it so that they actively operate with varying levels of decision-making and execution power. This creates novel risks and requires rethinking insider threat models (as these AI agents can act independently, make decisions, and even bypass traditional security controls). As agency is a core component of AI agents, their behavior must be continuously monitored; potentially even more rigorously than human users -to detect misuse, unintended actions, and emerging risks.

Partnering for Impact: Zenity + OpenPolicy

To amplify our impact, Zenity has partnered with OpenPolicy and are excited to participate in this week’s Fly-In in Washington D.C.. Through this partnership, we’re actively contributing to written guidance, engaging in policy discussions, and serving as a technical resource for regulatory bodies. As standards and expectations solidify, Zenity will remain at the forefront, helping to ensure that this new era of AI is built on secure, responsible foundations.