Key points
- The security focus has broadened from traditional code vulnerabilities to risks involving agent autonomy.
- The session connects past, current, and future AI attack surfaces.
- Defense is presented as a practical concern for systems using increasingly autonomous AI.
Context
This item summarizes an SDC2026 session on the shifting security threats linked to large language models and agents. For tech teams, the discussion positions AI security as a system-wide concern that spans code, agent behavior, and architecture.
What this means for you
Directly applicable
It gives developers and technical decision-makers a framework for reviewing AI security risks across code, agent behavior, and system boundaries.
Who benefits: Developers and technical decision-makers can use the review to sharpen security reviews of AI-enabled and agentic systems.
This site can provide architecture consulting and legacy-system analysis for teams evaluating how AI security concerns affect their software landscape.
Share article: