Bottom line: Against AI-powered attacks, a preventive strategy is recommended, using self-controlled defensive AI that stops data exfiltration and lateral movement before external security filters could block one’s own response in an emergency.
In light of AI-powered attacks, security experts are calling for a strategic shift toward preventive measures. At the center of this is defensive AI that companies operate and control themselves, in order to halt data exfiltration and lateral movement at an early stage.
The approach described here treats prevention as the overriding principle for dealing with AI-powered attacks. Organizations are meant to stop data exfiltration and lateral movement within the IT infrastructure before it develops into a full-blown security breach. A central element of this strategy is defensive AI that is operated and controlled by the company itself – in contrast to externally provided solutions, where a security filter could end up blocking the organization’s own defenses in a real incident.
For CISOs, this translates into a concrete requirement for the architecture of their own security infrastructure: defense mechanisms must not depend on third-party providers whose control mechanisms could restrict the organization’s own response capability at the moment of an active incident. If, for example, an AI-powered attack method attempts to break out of an isolated environment (“sandbox”), the defense must be able to take effect in real time and without external dependencies.
Shifting the focus from reactive detection measures to preventive controls means that security leaders need to review existing processes for monitoring data movement and internal network activity. In particular, the question of who operates the defensive AI being used, and how much operational control the company itself retains over it, should factor into the evaluation of security solutions.
Source: itwelt.at · Published August 4, 2026
Lumi AI News — AI-assisted curation pursuant to Art. 50 EU AI Act. Paraphrasing and classification by Lumi News Pipeline v1.8.3.