Toward an Ontology of Adversarial Information: Cybersecurity and the Ontological Limits of Floridi's Veridicality Thesis

Luciano Floridi's Veridicality Thesis (VT) holds that data constitute semantic information only if true; false data are therefore pseudo-information. This paper argues that the VT was built for semantically cooperative environments and becomes ontologically insufficient in adversarial sociotechnical contexts, where agents deliberately produce structurally valid, semantically functional, and causally efficacious, yet false, data. The central insight is that computational systems do not operate on truth — they operate on architectural operational validity — a distinction the classical Philosophy of Information has not adequately theorized. We propose a Relational-Operational Ontology of Adversarial Information (ROOAI) formalized around three components: (i) the interpretive system triplet ⟨Π, V, Δ⟩, comprising protocolar repertoire, validation mechanism, and state-update function; (ii) formal definition D1 with four constitutive conditions; (iii) six hierarchical pillars centered on operational validity. ROOAI does not deny the epistemological relevance of truth — it denies its universal ontological sufficiency in adversarial environments. A three-type taxonomy (Forged, Corrupted, Injected) and five formal properties are derived. Five objections are raised and answered. The model is extended to generative adversarial AI — prompt injection, data poisoning, adversarial examples. Legal implications are examined strictly within the bounds of the ontological grounding. The paper concludes that semantically hostile environments require an ontological extension of classical Philosophy of Information, opening a research program in adversarial information ontology.