Cybersecurity firm Resecurity has flagged an uptick in reconnaissance activity by foreign adversaries probing national AI infrastructure worldwide, aiming to map exposed AI and machine learning systems back to specific organizations before exploiting misconfigurations to breach them.
Unlike traditional IT systems, AI environments leak operational intelligence through APIs, metrics dashboards, and orchestration tooling that most security teams aren’t yet monitoring closely.
Resecurity points to six categories under sustained attacker interest: inference servers, vector databases, model registries, orchestration platforms, notebook environments, and GPU-backed compute.
Because these systems are tightly interconnected, a single weak point such an exposed notebook or an unsecured dashboard can expose an organization’s entire machine learning ecosystem at once.
Motivations diverge by actor: cybercriminal groups are largely after data to monetize or ransomware to deploy, while nation-state and advanced cyberespionage operators are focused on intellectual property theft.
Resecurity also warns of AI-specific attack paths, including GPU hijacking for unauthorized workloads, supply chain poisoning through tampered models, and manipulation of vector databases used in retrieval-augmented generation systems.
The firm cautions that the complexity of modern AI stacks is already slowing incident response, and that successful breaches can expose entire training datasets and operator query logs containing confidential information.
For a deeper technical breakdown of how attackers fingerprint and enumerate AI infrastructure — including specific ports, APIs, and mitigation steps — see our extended coverage on Military AI.
