Research journeys · Record 003
Making AI-Assisted Claims Independently Challengeable: Publication Authority and a Protocol for Falsifiable Publication Records
How can AI-assisted claims remain connected to evidence, analysis, human authorisation, the presented surface and later corrections? This public preprint develops and examines a protocol candidate for that question.
- Status
- Preprint · not peer reviewed
- Version
- v2 · revised 21 October 2026
- Subjects
- cs.AI · cs.CR · cs.CY
Chapters
The question behind it
A publication should remain independently challengeable.
The preprint starts from a problem: AI-assisted claims can appear authoritative when evidence, analysis, human authorisation, presentation and correction history refer to different states.
The proposed permission to publish (Publication Authority) applies to one exact checked state: it cannot be transferred and can only be used once. The preprint sets out its rules as a machine-readable protocol candidate called the Publication-Accountability Calculus (PAC-2026).
The rules and accompanying models can be checked by machine. That is what ‘executable publication semantics’ means here, not a running end-to-end system.
Understand the approachSix bound obligations
The fourth preserved version of the checking rules is called SF-4 (semantic freeze) in the preprint. It joins six obligations: passing one checkpoint cannot compensate for missing evidence at another. This list summarises the source, not an established standard.
- 01
Evidence
Evidence is bound to the state being examined.
- 02
Runs & artefacts
Runs and artefacts remain part of the check.
- 03
Measurement
Measurement and its disclosure are included.
- 04
Authorisation
Authorisation remains a distinct, human-bound check.
- 05
Surface correspondence
The presented surface must correspond to the accepted record.
- 06
Lifecycle continuity
Correction and lifecycle information remain connected.
What the preprint reports
Bounded internal evidence, not a universal result.
Selected reported results from public version v2: section 5.3 covers the finite models, section 5.6 the internal implementation tests, and Supplement 8 the rerun of archived observations. Nothing was rerun or independently recalculated for this record.
Explore the reported internal results
- finite formal models (not AI language models)
- 10
- reachable states without violations of the checked safety rules
- 110,764
- unsafe configurations producing the expected violation or countermodel
- 76
- matched scored expectations in the in-house test after shared integration
- 183
- archived observations rerun on the same host; not additional scored cases
- 240
‘Safe’ applies only within the checked models. Case classes were known; individual instances were not known at the freeze. In-house agreement followed shared integration: 21 confirmations of known baseline cases are not independent new findings. For 36 scored instances, the Rust terminal classifications were bridge-derived rather than native confirmations. The rerun retains a scored denominator of 183. The 240 observations are neither another blind experiment nor external replication.
Where to go from here
The specification follows the reference model.
The preprint is its own public research state. It remains distinct from the published BIS reference-model paper, software artefacts and any later independent or field-oriented validation.
- Explore the BIS reference model
The published Springer contribution with its own evidence and status.
- Research on TechJournal
Further research journeys and their bounded sources.
- AIJIM Protocol website
The protocol website. The preprint version cited here remains separately linked on arXiv.
Read the original
Preprint and research artifacts.
The current version is v2, revised 21 October 2026, first submitted 15 September 2026; not peer reviewed. The new evidence collection and the historical September artifacts remain separate sources, not additional articles.
- 01
Preprint · not peer reviewed
Making AI-Assisted Claims Independently Challengeable: Publication Authority and a Protocol for Falsifiable Publication Records
arXiv:2609.17631v2 · cs.AI, cs.CR, cs.CY · all eight authors affiliated with UTS
Yifei Dong is corresponding author · v2 · revised 21 October 2026 · first submitted 15 September 2026 · DOI 10.48550/arXiv.2609.17631
Read the preprint on arXiv - 02
Separate evidence collection · Software / research artifact
AIJIM Publication Authority — Executable Evidence for Claim-Edit Regions and Publication Protection
v1.0.0 · 18 October 2026 · DOI 10.5281/zenodo.23450434
Executable research artifacts for claim edits and publication protection. A separate collection, not another article or a new normative protocol version; not evidence of independent validation.
Archived artifact on Zenodo - 03
Historical SF-4 archive · Software / research artifact
AIJIM Publication Authority — Specification and Reproducibility Artifacts
v0.1.0 · 12 September 2026 · DOI 10.5281/zenodo.22727750
Archive of the evaluated SF-4 specification state. Retained as an exact source; the later evidence collection does not replace this historical object.
Archived artifact on Zenodo - 04
Separate correction candidate · Software / research artifact
AIJIM Publication Authority — Specification and Reproducibility Artifacts
v0.2.0 · 15 September 2026 · DOI 10.5281/zenodo.22763325
Correction candidate within the original September collection. Does not retroactively change the evaluated SF-4 state and is not another paper publication.
Archived artifact on Zenodo
Cite this work
Current version v2 is available with its revision date, exact arXiv URL, DOI and preprint status. The exports retain the first-submission date in a note.
Take another look
A preprint is a visible interim state.
The current reading link leads to v2; the preprint remains not peer reviewed. The first contextual record from 22 September concerned v1 and is preserved as history. Today’s source synchronization is not a new independent assessment of the research.
- Which objections would change the specification or its boundaries?
- What evidence would be needed for independent reproduction?
- Which validation remains outside this preprint?
Public arXiv metadata, abstract, version and status linked. Reported internal results kept distinct from general efficacy or truth claims.
Linked the public revision of 21 October and tied the existing figures to exact full-text locations. Added the separate evidence collection of 18 October; retained the SF-4 archive and correction candidate. Updated citation exports to v2 and preserved the original v1 exports separately. No reproduction or new validation performed.
The work remains challengeable.
Further research records show what each source supports — and what it leaves open.
Explore the sources