Bottom line: According to Netskope’s AI Risk and Readiness Report, AI risk arises simultaneously on four levels – employees, developers, in-house models and agents – which is why isolated measures are not enough to achieve real AI security.
In its latest AI Risk and Readiness Report, Netskope warns that AI usage in enterprises now takes place simultaneously on four levels – employees, developers, in-house models and autonomous agents. Anyone who secures only one level leaves the other three unprotected.
Netskope’s AI Risk and Readiness Report shows that AI in enterprises can no longer be treated as a single, clearly delineated tool. Instead, risk arises in parallel across four different levels: through employees using AI applications in their daily work, through developers integrating AI functions into their own software, through self-trained or fine-tuned models, and through autonomous AI agents that carry out actions independently. The report condenses this multi-layered reality into five concrete use cases which, according to Netskope, reveal an organization’s actual security posture.
For CISOs, this finding means that classic, isolated protective measures fall short. A security strategy that focuses solely on controlling employee access to public AI services, for example, overlooks the risks arising from self-developed models or autonomously acting agents in development environments. Because these four levels are often the responsibility of different teams – IT security, software development, data science departments and business units – there is a high risk of governance blind spots if responsibilities are not coordinated across functions.
For practical implementation, this creates the need to establish visibility and control across all four levels rather than relying on isolated individual measures. CISOs should check whether existing monitoring and governance processes actually capture employee usage, development activities, internal models and agent behavior equally. The report provides five concrete use cases as a starting point for taking stock of an organization’s own AI security posture.
Source: www.security-insider.de · Published August 14, 2026
Lumi AI News — AI-assisted curation pursuant to Art. 50 EU AI Act. Paraphrasing and classification by Lumi News Pipeline v1.8.3.