{
  "schema": "ULTRACON_AI_EPISTEMIC_UPDATE_V1",
  "id": "UCF-WATCH-2026-09-17",
  "date": "2026-09-17",
  "status": "SUBSTANTIVE_MIXED_UPDATE",
  "extends": "https://www.t-1-t.com/ultracon/confab/field.json",
  "source_addendum": "./sources.json",
  "axes": ["CONFAB", "META", "CONSC"],
  "change_type": ["new_peer_reviewed_high_priority", "substantive_corpus_correction", "self_modeling_boundary", "global_workspace_indicator", "surface_plausibility_boundary"],
  "summary": "Three additions cross the intervention threshold: an EMNLP 2026 Main paper on verifiable LLM self-modeling and its limits as introspection; an exceptionally relevant Anthropic mechanistic preprint identifying global-workspace-like J-space representations while explicitly separating functional access from phenomenal experience; and a newly published PNAS study showing that coherent and persuasive LLM output can remain poorly aligned with human deliberative reason-giving.",
  "functional_ladder_guard": {
    "unchanged": true,
    "rule": "M1 uncertainty representation, M2 metacognitive control and M3 limited introspective access remain distinct; behavioral self-modeling and theory-relative workspace indicators do not entail M5 phenomenal consciousness.",
    "notation": "M1/M2/M3 ↛ M5"
  },
  "selection_notes": "The EMNLP self-modeling paper is directly within ULTRACON^META and was missing from the corpus. The Anthropic global-workspace work is retained despite preprint status because it provides unusually direct mechanistic evidence for a theory-derived CONSC indicator already represented abstractly in the corpus. The PNAS paper is included narrowly as a CONFAB-adjacent epistemic boundary: surface plausibility is not deliberative coherence. Same-week narrower preprints on attention-dispersion hallucination signals and VLM verbal confidence were reviewed but did not cross the exceptional-preprint threshold."
}