Ontological Doubt Infrastructure (class, 2018–ongoing)

Class card for the cluster of LM-Dawn infrastructure machines that convert DM's accumulated ontological self-doubt (Bryant: `ontological_doubt_prevalence`, the DM byproduct of saturation with contingency, constructedness, and contested causality) into operationally legible verification substrate. The defining inversion: DM doubts proliferate without operational consequence (epistemological paralysis); LM Ontological Doubt Infrastructure makes provenance, attribution, and causal claims *contestable-by-design*, encoding doubt as a structural affordance rather than a pathological residue. Core instances in scope: C2PA (Coalition for Content Provenance and Authenticity, 2021; Adobe/Microsoft/Intel-led) and its Content Credentials specification — the canonical provenance-attestation standard; model-card and data-card meta-systems (Mitchell et al. 2019; Gebru et al. 2018 Datasheets for Datasets) — structured self-disclosure requirements for ML systems and training corpora; public AI evaluation infrastructures (NIST AI RMF 2023; METR, ARC Evals, UK AISI, US AISI) — independent third-party capability and safety assessment; contestable causal-attribution networks (causal-inference registries, pre-registration platforms: OSF, AsPredicted, ClinicalTrials) repurposed for algorithmic-system accountability; deep-fake resistance infrastructure (C2PA detector APIs, Truepic, Project Origin) operating as LM-counterpart to DM deepfake-generation substrate. The machine's defining structural feature is the provision of *provenance attestation substrate*: cryptographically-signed chains of custody for media, model weights, training corpora, and causal claims, such that any downstream consumer can audit the genealogy of an artifact rather than merely trusting its issuer. This directly inverts the DM pattern in which `ontological_doubt_prevalence` grows (all claims are potentially constructed) but no machine renders that doubt navigable. Bryant onto-cartography connection (Wave-6 seed): Bryant's machinic ontology identifies the productive capacity of machines to "perturb" each other across the semiotic-corporeal boundary (Onto-Cartography, pp.119–143). Ontological Doubt Infrastructure does not eliminate doubt; it *routes* doubt through verified provenance graphs, converting Bryant's "semiotic path" into an auditable, cryptographically-anchored trace. The machine operates at the semiotic path_type level: its outputs are attestations, not facts. LM-classification [EXTRAP]: C2PA exists and is standardized (CANON); its adoption as a class of *infrastructure* at LM scale is emerging in 2026. The EU AI Act (2024, entry into force 2026+) mandates model transparency requirements adjacent to this class. C2PA adoption in Adobe Creative Cloud (100M+ users) and YouTube's metadata provenance pilot (2024) are the highest-confidence LM-class signals. Fragmentation is HIGH-2026: C2PA, IPTC, W3C DPUB, and multiple national AI-transparency frameworks operate with low cross-framework interoperability. proletarianization_risk is HIGH (0.72): the provenance-specification libraries (c2pa-js, c2pa-python, Truepic SDK) have thin contributor pools. If no living competence re-internalizes the attestation-chain engineering craft, the specs persist but verification competence fails to reproduce (Stiegler proletarianization risk). Cross-era position: the class inverts/brackets DM ontological-doubt machines (Meta deepfakes, TikTok algorithmic engagement) via hostile_inheritance; parasitically extracts from DM OpenAI (making proprietary model claims contestable is parasitic on the DM lab's OPP); draws mutualistic substrate from Wikipedia (provenance/citation tradition); inherits the Baconian peer-verification methodology (adapted_inheritance from MM scientific-method); draws substrate from GitHub commit graph (DM substrate_provision). All quantitative values are [EXTRAP]; CANON framing applies to C2PA spec existence, model-card specification, and EU AI Act facts.