00:41 13-02-2026

NATO Warns of AI Model Poisoning in Information War

RusPhotoBank

NATO analysts warn that deliberate AI model poisoning with false data could become a key tool of information warfare as the Pentagon eyes civilian AI.

Analysts at NATO’s Strategic Communications Centre of Excellence have released a report titled «Next-Generation Information Environment».

In their assessment, experts argue that deliberate «poisoning» of artificial intelligence models with false or distorted data may become one of the tools of modern information warfare. The report describes a «new front» in this struggle as the intentional corruption of Western open-source AI models by feeding them fabricated information designed to warp their future training cycles.

The authors believe that, going forward, AI systems could turn into some of the most influential participants in information processes. At the same time, they warn that large-scale data manipulation and targeted «poisoning» of models by adversaries risk eroding public trust in these technologies.

Earlier, Reuters reported that OpenAI, Anthropic, Google and xAI have been in talks with the Pentagon about deploying their neural networks in closed segments of US military infrastructure. According to these reports, the US Department of Defense is interested in using civilian AI technologies without the standard restrictions that limit their application in combat conditions.

Reuters also noted that during a meeting at the White House, Pentagon Chief Technology Officer Emil Michael signalled to company representatives that the military intends to integrate AI models into both open and classified domains of its own networks.