One does not simply defend agentically
NCSC UK argues that agentic AI defense is limited by organizational politics and proposes a risk framework for safe automation.
The NCSC UK discusses the challenges of using agentic AI for cyber defense, arguing that defensive problems are often organizational rather than purely technical. The post outlines a framework to assess the risk of automated defensive actions, emphasizing the need for human oversight and controlled scoping to avoid business disruption.
- Defensive problems are often political/organizational, not just technical.
- AI automation for defense is constrained by business risk and accountability.
- A framework is proposed to estimate the 'riskiness' of automated defensive actions.
- Defenders need explicit risk frameworks to safely use AI automation.
- Countries
- UK
From the original article
7 min read at ncsc.gov.uk
Defenders can’t use AI in the same way attackers can, but there’s much they can do to unlock the potential of agentic cyber defence.
ncsc.gov.uk
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.