Gravitee Sets a Framework for AI Agent Governance
AI Agent Governance Needs Clear Accountability
AI agent governance is becoming urgent as enterprises move from experiments to production systems. In this Techstrong TV conversation, Mike Vizard talks with Rory Blundell, CEO of Gravitee, about a framework for securing AI agents before they create business risk. The discussion centers on a simple idea. Agents need identity, role, authority, oversight and recourse before organizations can trust them at scale.
Blundell explains that AI agents are not just another automation layer. They can act at machine speed, reach into systems and make decisions that affect data, operations and customers. That changes the security conversation. Teams need to know who owns each agent, what the agent is allowed to do and what happens if something goes wrong.
Trust Becomes the Foundation for AI Value
Gravitee’s view is that AI agent governance should enable adoption rather than slow it down. Blundell says many CIOs and CTOs are under pressure to put agents into production. Yet the transcript notes that only about 20% of those production agents are considered fully secured by the leaders deploying them.
That gap can limit the value of AI. One incident can wipe out months of expected return from an agent project. A trust layer gives teams a way to keep moving while reducing the attack surface. It also helps business leaders understand which agents exist, what services they touch and where human responsibility sits.
Old Security Lessons Still Apply
The conversation also makes the case for applying proven security thinking to new AI systems. Blundell argues that organizations should not forget the governance, security and control models they built during the internet era. APIs, MCP services, event streams, LLMs and AI agents all need abstracted control layers.
That approach does not require a single silver bullet. It requires defense in depth. Guardian agents, identity controls, service boundaries and human oversight can all help reduce risk. The goal is not to stop AI adoption. The goal is to make adoption safer and more sustainable.
InfoSec Has a New Role in Human-Agent Collaboration
For security leaders, the episode offers a practical message. They do not need to become the people who simply say no to AI. Instead, they can help build the conditions for human-agent harmony. Blundell argues that people and agents working together can create more value than either side alone.
That value depends on trust. If employees do not trust agents, they will resist using them or avoid giving them meaningful work. Strong AI agent governance gives organizations a path to productivity without giving up control. It also gives security teams a central role in turning agentic AI from a risky experiment into an operating model.
