Securing the Agentic Enterprise from Red Team to Runtime | BTF 2026
- Mar 18
- 2 min read
Updated: Apr 6
On March 13, ARMIS took part in Building The Future 2026 with a high impact technical session focused on one of the most critical challenges organizations face today securing AI agents in enterprise environments.
Titled Securing the Agentic Enterprise From Red Team to Runtime, the session was led by Carlos Faria and Vitor Pinho and delivered a clear and practical overview through a Live Demo on the security risks introduced by AI agent adoption and how enterprises can address them with confidence.
AI agents as a new attack surface
The session highlighted a fundamental shift AI agents do not behave like traditional software. By operating through natural language, contextual reasoning, autonomy and external data sources, they introduce new attack surfaces that are not covered by conventional application security approaches.
A central topic was prompt injection, addressed as a structural characteristic of large language models rather than a simple vulnerability. When models are unable to distinguish between data and instructions, the risk of manipulation increases through both direct interactions and indirect inputs such as documents or external knowledge sources.
From red team to runtime Security as a layered strategy
ARMIS reinforced a key principle security is not a feature but a layered strategy. Protecting AI agents requires an end to end approach that spans design, testing, deployment and runtime operations.
Key takeaways included proactive red teaming of AI systems before deployment, defense in depth through multiple security layers and the integration of AI related threats into SOC monitoring and response processes. This approach enables organizations to adopt AI safely while preserving existing architectures.
Real world scenarios with real risks
To ground the discussion, the session presented a real example of an AI powered hiring agent capable of analyzing resumes, extracting skills and interacting naturally with candidates. While highly effective from a user experience perspective, the example demonstrated how such systems can quickly become high value targets if security is not embedded from the start.
The message was clear greater autonomy requires greater responsibility in securing AI agents.
Security for agentic AI is available today
One of the strongest messages of the session was that these security capabilities are not future concepts. They are available today and can be adopted incrementally without the need to redesign existing applications.
Securing AI agents is not a long term roadmap item. It is an actionable priority.
ARMIS leading secure AI adoption
ARMIS participation in Building The Future 2026 reinforces its commitment to shaping responsible and secure enterprise AI adoption. By combining offensive security, engineering expertise and operational readiness, ARMIS continues to help organizations innovate with confidence in an increasingly agent driven world.
Watch the video at Building the Future official youtube channel:
