Agentic AI
,
Artificial Intelligence & Machine Learning
,
Cloud Security
Cloudflare’s Christian Reilly on Securing AI Agents Before They Scale
Agentic artificial intelligence has outpaced the security frameworks organizations built to govern it. The same availability that makes agent technology powerful is what also makes it dangerous. Security teams have no visibility into agents they didn’t sanction, and that blind spot is growing, said Christian Reilly, field CTO at Cloudflare.
See Also: Edge Transformation: Top 5 SASE Predictions and Trends
Organizations don’t need to slow adoption. Instead, they need to establish identity, access and operational controls before agent deployments scale. Without those guardrails, security teams risk creating blind spots that expose organizations to unintended actions, unauthorized access and emerging forms of shadow AI.
“My biggest fear is that there’s lots of agents running around that the security parts of organizations have absolutely zero visibility into, and therefore not really sure about the risk of those agents in operation,” Reilly said.
In this video interview with ISMG at Infosecurity Europe 2026, Reilly also discussed:
- The risks of shadow agents and unmanaged Model Context Protocol deployments;
- Why prompt and response controls are critical for customer-facing AI applications;
- How regulated industries can apply human-in-the-loop accountability to agent workflows.
Reilly works with enterprise clients, prospects and partners to demonstrate the value of Cloudflare’s global network, collaborating across engineering, product and go-to-market teams. Prior to Cloudflare, he was CTO for EMEA at HashiCorp, where he focused on multi-cloud infrastructure and security life cycle management strategies.
Click Here For The Original Source.
