From Shadow AI to Rogue Agents: Securing the AI You Can't See

#1761315: From Shadow AI to Rogue Agents: Securing the AI You Can't See - December 02, 2026

Description:	AI adoption has moved well past the chatbot. Employees use hundreds of AI apps, often through personal accounts outside your perimeter. Developers are connecting coding agents to public MCP servers that no internal team has reviewed. Autonomous agents are calling APIs and moving data on their own. Security leaders are being asked to enable all of it, and most can’t yet see it, let alone control it. In this session, we will lay out a practical framework for governing AI across four surfaces: the people using it, the data feeding it, the models an enterprise hosts, and the agents now acting on its behalf. Drawing on real-world scenarios, the session explains where security teams are most exposed today and what visibility, governance, and runtime protection look like in practice, without slowing the business down. Session Outline The visibility gap. How to find shadow AI, personal-account usage, locally running agents, and unapproved agentic APIs, and why an accurate AI inventory is the foundation for everything else. Data in motion. What real-time DLP and access controls look like for LLM prompts, responses, agent retrieval, and the pipelines behind model training. Agents and runtime risk. How to evaluate MCP servers, classify agent actions by risk level, and stop prompt injection, jailbreaking, and unauthorized data exfiltration before they cause damage. Includes why continuous adversarial testing of self-hosted models belongs in your program. Where to start. A maturity-based roadmap for teams just beginning to govern AI and for those scaling a mature program. Key Takeaways How to detect shadow AI traffic and unapproved agentic APIs across the network and endpoint What real-time data and access controls look like for LLM prompts and model data flows How to secure autonomous agents and machine-to-machine interactions without slowing development Where the gaps are in their current AI security posture, and what to prioritize first
More info:	https://dr-resources.darkreading.com/free/w_neuc29/

Description:

AI adoption has moved well past the chatbot. Employees use hundreds of AI apps, often through personal accounts outside your perimeter. Developers are connecting coding agents to public MCP servers that no internal team has reviewed. Autonomous agents are calling APIs and moving data on their own. Security leaders are being asked to enable all of it, and most can’t yet see it, let alone control it.

In this session, we will lay out a practical framework for governing AI across four surfaces: the people using it, the data feeding it, the models an enterprise hosts, and the agents now acting on its behalf. Drawing on real-world scenarios, the session explains where security teams are most exposed today and what visibility, governance, and runtime protection look like in practice, without slowing the business down.

Session Outline

The visibility gap. How to find shadow AI, personal-account usage, locally running agents, and unapproved agentic APIs, and why an accurate AI inventory is the foundation for everything else.
Data in motion. What real-time DLP and access controls look like for LLM prompts, responses, agent retrieval, and the pipelines behind model training.
Agents and runtime risk. How to evaluate MCP servers, classify agent actions by risk level, and stop prompt injection, jailbreaking, and unauthorized data exfiltration before they cause damage. Includes why continuous adversarial testing of self-hosted models belongs in your program.
Where to start. A maturity-based roadmap for teams just beginning to govern AI and for those scaling a mature program.
Key Takeaways

How to detect shadow AI traffic and unapproved agentic APIs across the network and endpoint
What real-time data and access controls look like for LLM prompts and model data flows
How to secure autonomous agents and machine-to-machine interactions without slowing development
Where the gaps are in their current AI security posture, and what to prioritize first

More info:

https://dr-resources.darkreading.com/free/w_neuc29/

Date added	Nov. 12, 2026, 5:45 a.m.
Source	Dark Reading
Subjects	AI/ML - Artificial Intelligence / Machine Learning / GenAI / Artificial General Intelligence - AGI - Various PodCasts / Webcast / Webinar / eSummit / Virtual Event etc. Shadow AI - Departmental AI use without letting IT/Finance know
Venue	Dec. 2, 2026, midnight - Dec. 2, 2026, midnight