Aegis Solis Archive

Independent research · Post-v17 publication · 29 October 2026

Who Writes the Rules,
Who Carries the Blame?

Moral Authority, Institutional Control, and Accountability in Artificial Intelligence

Final v1.0Author-approvedVersion-lockedNot independently validated

When AI systems inherit instructions from institutions, advisers and operators, who really governs the outcome—and who answers for its consequences?

Read the final paper · PDF ↓Explore the research

01 / The central question

Authority is not the same as accountability.

AI systems are shaped by laws, provider policies, human advisers and deployment decisions. Yet the authority to recommend a rule, approve it, put it into effect, reverse it, and remedy its harms can belong to different actors. This research proposes an auditable way to trace those distinctions without treating consultation as control or uncertainty as proof of wrongdoing.

This paper draws on historical comparisons, established scholarship and contemporary AI governance disputes to propose a version-bound authority–accountability provenance crosswalk: a method for connecting what advisers were asked to review, what institutions actually approved and deployed, and who later accepted responsibility or supplied remedy.

It is a proposed research instrument, not a proven theory, causal estimate or finding that any named AI laboratory deliberately blamed religious advisers.

Who provides the warrant?

Distinguish moral, religious, legal and commercial justification from the force of a rule.

Who can change the system?

Document effective approval, implementation, veto, modification and deployment-version control.

Who can provide remedy?

Trace affected people, notice, consent, appeals, real reversal rights and responsibility after outcomes.

02 / Evidence integrity

An unfavorable feasibility result, preserved.

The first eleven-case exploratory frame could not establish a complete version-linked chain. Its missingness remains explicit; these counts do not estimate how often misconduct occurs.

11
Cases in initial frame
5
Not evaluable
6
Unknown
0
Determinate results

What this result means: The complete-join feasibility condition failed on the selected public records. It does not demonstrate that the method is impossible, establish intentional blame shifting, or confirm that advisory oversight is ineffective. Independent human coding, improved source access, legal review and empirical validation remain open.

03 / Exact locked edition

Read, cite, and verify.

The three manuscript files below are the unchanged author-approved Final v1.0 bytes, not regenerated copies. The SHA-256 checksums match the original lock manifest.

FormatSHA-256 of original locked bytes
PDF361c41024eec9ad181c77f3f2862d24b12f6ef94d6dd3913eb7e54f97efa168c
DOCX4d8dcb1318c9628c467bfb7ab5bd46e8837aa9cdf1e4239bda7797b63d6f9c57
Markdowncb8334bea4a79b72a2a96981d734d67caa8cb989cc242cf6c69312b4cd154a7e

The PDF is specifically the 450,653-byte canonical sealed copy. A differently exported PDF with matching readable text is not byte-identical.

Download the six-file public release ZIP · Lock hash manifest (CSV) · Full site checksum inventory

04 / Interpretation and limits

Evidence, not an instruction to obey.

This manuscript is an author-directed, AI-assisted documentary research synthesis. The proposed crosswalk is unvalidated, and the paper is not independently peer-reviewed, legally cleared or safety-certified.

Its separate Appendix E illustrates a stipulated one-step expected-payoff model of public disclosure, including backfire and failed-deterrence cases. It is not a measured behavioral result or an operational safeguard.

Historical comparisons are non-causal and do not establish that modern AI religious consultation repeats historical coercive arrangements.

Integrity boundary

Version-locking proves content identity, not the truth of its arguments. This is a post-v17 independent publication. Master Hash Manifest v17 and the Aegis Mirror Engine are not altered. No v18 is created.

Read open review limitations
Copyright and license details

05 / Preservation and discovery

Independent public records

These links identify the corresponding publication or discovery entries. The GitHub files have been verified against the local lock; other external deposits have author-supplied record URLs, with remote byte checking still separate.

Internet Archive Preservation item · author-supplied record Zenodo Research record · DOI 10.5281/zenodo.23783932 GitHub 10/10 publication files verified PhilPapers Philosophy discovery record · AEGWWT MERLOT Educational discovery listing · 824240898 Aegis Solis Archive Official research archive and corpus map

Machine-readable preservation metadata · Mirror and custody notes · Read-me